Indole peroxygenase activity of indoleamine 2,3-dioxygenase

Abstract

The heme enzyme indoleamine 2,3-dioxygenase (IDO) was found to catalyze the oxidation of indole by H₂O₂, with generation of 2- and 3-oxoindole as the major products. This reaction occurred in the absence of O₂ and reducing agents and was not inhibited by superoxide dismutase or hydroxyl radical scavengers, although it was strongly inhibited by l-Trp. The stoichiometry of the reaction indicated a one-to-one correspondence for the consumption of indole and H₂O₂. The ¹⁸O-labeling experiments indicated that the oxygen incorporated into the monooxygenated products was derived almost exclusively from H₂¹⁸O₂, suggesting that electron transfer was coupled to the transfer of oxygen from a ferryl intermediate of IDO. These results demonstrate that IDO oxidizes indole by means of a previously unrecognized peroxygenase activity. We conclude that IDO inserts oxygen into indole in a reaction that is mechanistically analogous to the “peroxide shunt” pathway of cytochrome P450.

The heme enzyme indoleamine 2,3-dioxygenase (IDO) catalyzes the first and rate-determining step of l-tryptophan metabolism in nonhepatic mammalian tissues by inserting both atoms of O₂ into the indole ring to form N-formylkynurenine (N-FK) (1). This dioxygenase activity of IDO requires O₂ and the reduced (Fe²⁺) enzyme or O₂^•− and the oxidized (Fe³⁺) enzyme. In recent years, an increasing number of physiological consequences of IDO activity have been identified, including links to neural, ocular, and immunological disorders (2). Of these roles, the ability of IDO expressed by tumors to suppress the normal response of T lymphocytes and enable tumor survival and growth has attracted the most intense interest, owing to its implications for the development of new therapeutic approaches in cancer treatment (3).

Aside from tryptophan (Trp), the dioxygenase activity of IDO extends to oxidation of other indoleamines, such as 5-hydroxy-l-Trp, serotonin, and tryptamine, whereas indole, 3-methylindole, and indoleacetic acid are not substrates (4–6). Although indole can apparently bind to the oxygenated enzyme (IDOFe³⁺–O₂^•−) or to the IDOFe²⁺–CO complex (7), autoxidation to IDOFe³⁺ is unaffected by the presence of indole, and oxidation of indole does not occur (5). Notably, IDO does not catalyze oxidation of l-Trp by H₂O₂ (4, 8, 9). Consequently, the reactivity of IDO with H₂O₂ received little attention for more than 30 years. During that time, IDO was noted to have peroxidase activity (6) and to catalyze the H₂O₂-supported N-demethylation of benzphetamine and hydroxylation of aniline (10), but these activities were not studied further.

Recently, reactivity of the enzyme with H₂O₂ has received considerable attention in light of spectroscopic detection of a compound II-like ferryl species (11, 12) and quantum mechanics/molecular mechanics simulations implicating a role for this intermediate in the catalytic cycle (13). Subsequently, the peroxidase activity of IDO has received renewed attention (9).

Although the source of the oxygen incorporated into the products of IDO-catalyzed aniline hydroxylation remains unclear, the limited literature that is available implies that IDO may be capable of processing alternative substrates in the presence of hydrogen peroxide, presumably owing to the relatively accessible nature of the active site (14). In the present work, we have evaluated the ability of IDO to catalyze the oxidation of indole and methylated derivatives of indole by H₂O₂, identified the products formed in these reactions, and identified the source of the oxygen that is incorporated into the reaction products. These results provide evidence that IDO exhibits peroxygenase activity similar to the so-called peroxide shunt of cytochromes P450 (P450).

Results

IDO-Catalyzed Indole Oxidation by H₂O₂.

Addition of H₂O₂ to a solution containing IDOFe³⁺ and indole resulted in decreased absorbance at 276 nm and corresponding increases at ∼233 nm, 390 nm, and 610 nm (Fig. 1A) that are consistent with consumption of indole. The lack of isosbestic points in this family of spectra indicated that multiple products and/or intermediates were formed during reaction. Accordingly, at least seven products were identified by HPLC, and a MeOH/H₂O-insoluble pigment was identified by TLC to be indigo blue (Fig. 1B and Tables S1 and S2). Only the peaks with retention times corresponding to Tris buffer and/or enzyme (7.5 min) and indole (35.5 min) were detected when either H₂O₂ or the enzyme were omitted or when the enzyme was thermally denatured before reaction.

Fig. 1.

IDO-catalyzed oxidation of indole by H₂O₂. (A) Electronic spectra following aerobic addition of H₂O₂ (1 mM) to IDOFe³⁺ (750 nM) and indole (220 μM). (B) HPLC analysis of reaction components (I–VII) after addition of indole (2 mM) to IDOFe³⁺ (200 μM) and H₂O₂ (4 mM). Data supporting identification of products are provided in SI Materials and Methods.

Stoichiometry of Indole Oxidation and Inactivation of IDO.

IDO-catalyzed oxidation of indole by H₂O₂ initiated by aerobic addition of equimolar amounts of both substrates was monitored fluorimetrically to determine the stoichiometry of the reaction (Fig. 2A). Under these conditions (substrates, 26–416 μM; [IDO], 10 μM), indole was consumed nearly quantitatively (∼95%), consistent with one-to-one consumption of the two cosubstrates. Some peroxide was presumably consumed in downstream formation of dioxygenated indoles. Addition of more H₂O₂ on completion of reaction led to oxidation of the remaining indole and confirmed that depletion of H₂O₂ accounts for premature termination of the reaction. At lower enzyme concentration (e.g., 1 μM), less indole was oxidized, and reaction could be reestablished only by addition of fresh enzyme; however, diminishing amounts of indole were oxidized after each addition (Fig. 2B). In addition, rapid decay of the Soret band was evident with addition of modest concentrations of H₂O₂ to the free enzyme (Fig. S1A); thus, we attribute the loss in activity at lower [IDO] to oxidative damage to the enzyme by peroxide.

Fig. 2.

Stoichiometry of the oxidation of indole catalyzed by IDO and H₂O₂. (A) [Indole] oxidized after addition of equal [H₂O₂], as monitored by fluorescence emission, at 1 and 10 μM IDOFe³⁺ (dashed and solid lines, respectively). (B) Stalling and restoration of the reaction of indole (422 μM) with H₂O₂ (1 mM) and multiple additions of IDOFe³⁺ (500 nmol each) as indicated by the arrows.

Dependence of the Distribution of Oxidation Products on H₂O₂ Concentration.

The distribution of indole oxidation products after reaction (1 min) as a function of [H₂O₂] was analyzed by HPLC (Fig. 3). With initial [indole] > [H₂O₂], the major products were 2- and 3-oxoindole in a ∼3:1 ratio (I and V, respectively; Fig. 1B), with trace amounts of o-formylaminobenzaldehyde (II). Indole-2,3-dione (III) and 3-hydroxy-2-oxoindole (IV) were detected only when [H₂O₂] > [indole]. The yield of these latter products increased with [H₂O₂], but formation of V was maximal when [indole] > [H₂O₂], whereas I formed maximally when [indole] ∼ 2[H₂O₂]. In contrast, the yield of dioxygenated derivatives (II, III, and IV) increased with [H₂O₂], as also observed for VII, although the latter product was not quantified. The pattern of product accumulation suggested involvement of additional oxidation reactions of the initial products driven by excess peroxide.

Fig. 3.

[H₂O₂]-dependence of product distribution. Reactant concentrations and conditions are the same as for Fig. 1B, except for [H₂O₂] (0.5–6 mM): ○, indole; ■, 2-oxoindole; □, 3-oxoindole; ▲, o-formylaminobenzaldehyde; △, indole-2,3-dione; ◇, 3-hydroxy-2-oxoindole. Reaction time: 60 s.

Oxidation of 2-Oxoindole (I) and 3-Oxoindole (V).

The IDO-catalyzed oxidations of I and V by H₂O₂ were examined to determine whether one or both of these monooxygenated products are intermediates in formation of the dioxygenated products. Aerobic reaction of IDO with HPLC-purified I and H₂O₂ (1 min) yielded the dioxygenated products III and IV (Fig. 4A). In contrast to indole, conditions resulting in quantitative oxidation of I could not be identified. The basis for this observation remains unclear, but it is consistent with the accumulation of I during IDO catalysis of indole oxidation by H₂O₂.

Fig. 4.

Oxidation of 2- and 3-oxoindole by IDO. HPLC analyses of products from incubation (60 s except where noted) of 2- and 3-oxoindole with IDO and H₂O₂. (A) Products from aerobic incubation of 2-oxoindole with IDOFe³⁺ and H₂O₂ (2, 0.1, and 2 mM, respectively; bold line) or with no additions (dashed line). (B) Products from anaerobic incubation of 3-oxoindole with IDOFe³⁺ and H₂O₂ (1, 0.05, and 1 mM, respectively; bold line) or with no additions (dashed line), and products from aerobic incubation (2 h) of 3-oxoindole with no additions (thin line). (Inset) (i and ii) Kinetics of anaerobic oxidation of 3-oxoindole (100 μM) by with IDOFe³⁺ (2.5 μM) and H₂O₂ (100 μM), or H₂O₂ alone, respectively. (iii) Kinetics of aerobic incubation (2 h) of 3-oxoindole with no additions.

Preparation of V was performed in situ by deesterification of 3-acetoxyindole with porcine liver esterase, and all work with this preparation before HPLC was performed anaerobically to prevent spontaneous oxidation of V to indigo blue (15, 16). Under anaerobic conditions, V was stable for hours if either H₂O₂ or IDOFe³⁺ were excluded from solution, but with both enzyme and peroxide present, it was oxidized rapidly (Fig. 4B, Inset), and indigo blue resulted. 3-Oxoindolenine (VI) was detected among the reaction products, but no significant dioxygenated indole derivatives were formed (Fig. 4B). Nevertheless, reaction of 3-oxoindole with O₂ (∼2 h) in the absence of enzyme did afford III as a by-product, as reported previously (16). This result suggests that at least two pathways may account for this product in the IDO-catalyzed oxidation of indole: enzymatic oxidation of I and nonenzymatic oxidation of V by O₂. Intriguingly, neither II nor VII was detected when the initial substrate was either I or V, so it is most likely that these products are produced by a different pathway than that resulting in III or IV.

Source of Oxygen in the IDO-Catalyzed Oxidation of Indoles by H₂O₂.

Anaerobic oxidation of indole by H₂O₂ catalyzed by IDO resulted in products similar to those observed under aerobic conditions (Table S3 and Fig. S2). These reactions were then performed with H₂¹⁸O₂ to determine whether this oxidant is a source of oxygen in the reaction (Tables 1–3). Electrospray ionization (ESI)-MS analysis of I and V produced under aerobic conditions (¹⁶O₂-saturated) revealed that both exhibit a new [M+H]⁺ ion of m/z 136 in addition to the previously observed [M+H]⁺ ion (m/z 134) that was consistent with the incorporation of one atom of ¹⁸O into the monooxygenated products. The percent incorporation of ¹⁸O into I and V reflected the ∼90% ¹⁸O enrichment of H₂¹⁸O₂ used in these experiments. This result unambiguously demonstrates that the oxidation of indole by IDO and H₂O₂ occurred with essentially stoichiometric insertion of oxygen from H₂O₂ into the monooxygenated products. No loss of the ¹⁸O label in I occurred after 3 h at room temperature, whereas V oxidized spontaneously in this time.

Table 1.

Incorporation of ¹⁸O into indole oxidation products I and V

		[M+H]+ ion relative abundance, %
Product	Oxidant	m/z 134	m/z 136	m/z 138	18O incorporated, %*
I	H216O2	100	0.8	—	—
	H218O2	8.3	100	0.4	92 [102]
V	H216O2	100	0.7	—	—
	H218O2	10.6	100	0.7	90 [100]

—, below detection limit.

*Numbers in brackets are normalized based on the enrichment level of H₂¹⁸O₂.

Table 3.

Incorporation of ¹⁸O in indole oxidation product III

		[M+H]+ ion relative abundance, %*			18O incorporated, %†
Product	Oxidant	m/z 148	m/z 150	m/z 152	16O18O	18O18O
III	H216O2	100	1.2	0.3	—	—
	H218O2	28 (66)	100 (100)	2.2 (2.9)	78 [87]	0

*Numbers in parentheses denote relative abundances from a repeat analysis after a 3-h incubation.

^†Numbers in brackets are normalized based on the enrichment level of H₂¹⁸O₂.

Table 2.

Incorporation of ¹⁸O in indole oxidation products II and IV

		[M+H]+ ion relative abundance, %*			18O incorporated, %†
Product	Oxidant	m/z 150	m/z 152	m/z 154	16O18O	18O18O
II	H216O2	100	1.0	—	—	—
	H218O2	20 (100)	100 (75)	58 (9.1)	56 [62]	32 [40]
IV	H216O2	100	0.9	—	—	—
	H218O2	7 (9)	100 (100)	45 (35)	65 [72]	30 [37]

—, below detection limit.

*Numbers in parentheses denote relative abundances from a repeat analysis after a 3-h incubation.

^†Numbers in brackets are normalized based on the enrichment level of H₂¹⁸O₂.

ESI-MS analysis of the dioxygenated product II obtained in the same aerobic reaction resulted in two new [M+H]⁺ ions (m/z 154 and 152) and trace amounts of the previously detected ion (m/z 150). The relative abundance of these ions corresponded to 56% of singly ¹⁸O-labeled product and 32% of doubly ¹⁸O-labeled product. Over ∼3 h, ¹⁶O/¹⁸O exchange was evident and resulted in nearly complete disappearance of the m/z 154 ion and dominance of the fully ¹⁶O species. Both the lower than theoretical level of ¹⁸O incorporation in II seen initially and the subsequent, time-dependent ¹⁶O/¹⁸O exchange were consistent with behavior reported for the related compound N-FK (17) that exchanges oxygen with H₂¹⁶O (18).

Incorporation of ¹⁸O into III and IV was also evident. However, whereas at most one of the two oxygen atoms in III originated from H₂¹⁸O₂ (78% incorporation of a single ¹⁸O atom), a second ¹⁸O was also found in IV (65% and 30% incorporation of one and two atoms of ¹⁸O, respectively). Time-dependent loss of the ¹⁸O label from both products was evident, but this loss was slower for IV than for II or III, with just one oxygen atom (presumably the C₃ oxygen) exchanging with H₂O over 3 h.

Methylated indoles were also readily oxidized by IDOFe³⁺ and H₂O₂ to products listed in Table S1. As with indole, these reactions were also performed with H₂¹⁸O₂ (Table S4–S7). The aerobic reaction of 3-methylindole and 2-methylindole with IDOFe³⁺ and H₂O₂ proceeded with stoichiometric (92%) and partial (75%) incorporation of ¹⁸O from H₂¹⁸O₂ into their respective monooxygenated products. In contrast, no ¹⁸O incorporation was evident in the monooxygenated derivative of 2,3-dimethylindole. Dioxygenated products were also produced in the oxidation of 3-methylindole and 2,3-dimethylindole, but the mechanism of their formation evidently differs from that of indole, because neither product incorporated ¹⁸O.

Kinetics of IDO-Catalyzed Indole Oxidation as Supported by H₂O₂.

In general, the use of low [IDO] (1 μM) in indole oxidation resulted in ∼100 turnovers before reaction stopped. Inactivation of IDO under these conditions presumably results from peroxidative damage to the protein. Consequently, the kinetics of indole oxidation by H₂O₂ catalyzed by IDO has been described in terms of the rate of indole consumption as monitored fluorometrically. The rate of disappearance of indole as a function of [H₂O₂] in the IDOFe³⁺-catalyzed reaction exhibited Michaelis–Menten behavior, with K_m ∼1.1 mM for peroxide ([indole], 25–300 μM) (Fig. S3 and Table 4). In contrast, corresponding measurements as a function of [indole] exhibited biphasic kinetic behavior with increasing concentrations of substrate. This kinetic pattern resembled the behavior of several P450-catalyzed reactions for which it has been suggested that the availability a second substrate within the active site results in continuously increasing velocity at high [substrate] (19–21). A multisite kinetic model is also consistent with the conclusion that indole could bind to both IDOFe³⁺ and the IDOFe³⁺–Trp complex (7). Therefore, the concentration dependence of the initial rates was fitted to a single substrate, two-site model (19),

Table 4.

Kinetic parameters of the IDO-catalyzed oxidation of indole

[H2O2]		[Indole]
kcat, s−1	Km, mM	kcat1, s−1	Km1, μM	kcat2/Km2, min−1 μM−1
2.6 ± 0.1*	0.90 ± 0.08*	1.3 ± 0.2§	9 ± 4§	0.96§
7.6 ± 0.7†	1.3 ± 0.2†	0.34 ± 0.07¶	3 ± 4¶	0.48¶
20 ± 3‡	1.2 ± 0.4‡	—||	—||	0.18||

—, values of k_cat1 and K_m1 were too small to obtain reliable estimates.

Error estimates reported as SE.

*Data obtained as a function of [H₂O₂] at 25 μM indole.

^†Data obtained as a function of [H₂O₂] at 90 μM indole.

^‡Data obtained as a function of [H₂O₂] at 300 μM indole.

^§Data obtained as a function of [indole] at 500 μM H₂O₂.

^¶Data obtained as a function of [indole] at 500 μM H₂O₂ and 2 μM l-Trp.

^||Data obtained as a function of [indole] at 500 μM H₂O₂ and 5 μM l-Trp.

where K_m1 and V_max1 are the kinetic constants for the first site and K_m2 and V_max2 are kinetic constants for the second site. This analysis resulted in an apparent k_cat1 of 1.3 ± 0.2 s⁻¹, K_m1 of 9 ± 4 μM, and k_cat2/K_m2 of 0.96 min⁻¹ μM⁻¹ at 500 μM H₂O₂ (Table 4). Saturation of the second site was not observed in the range of [indole] examined. As little as 5 μM l-Trp strongly inhibited the reaction and changed the kinetics from biphasic to monophasic as the result of near abolition of k_cat1, presumably owing to the high affinity of Trp for the ferryl enzyme [K_d = 0.3 μM for the high-affinity binding site (9)]. Nevertheless, even [l-Trp] as great as 30 μM was insufficient to completely inhibit activity. l-trypotophan was not oxidized by IDOFe³⁺ and H₂O₂ to any significant degree.

Enzyme Reduction in the Reaction with Indole and H₂O₂.

IDO-catalyzed oxidation of indole was monitored anaerobically and with substoichiometric [H₂O₂] (∼0.8 molar equivalent relative to IDOFe³⁺) to determine the oxidation state of IDO after turnover. Electronic spectra of the enzyme under these conditions revealed formation of a minor fraction of IDOFe²⁺ (∼12%) in ∼15 min. An accompanying increase in absorbance at 608 nm indicated that indigo blue accumulated during this time. At higher [H₂O₂] (5 molar equivalents), the fraction of IDOFe²⁺ formed increased to ∼50% (Fig. S1B). Immediately after addition of H₂O₂, [indole] decreased briefly, but then stopped changing even though IDO reduction continued. Introduction of catalase after H₂O₂ addition did not influence this decrease in the enzyme. These results argue against the involvement of a coupled process, but the anaerobic conversion of V to indigo blue by IDOFe³⁺ under anaerobic conditions in the absence of H₂O₂ indicates that reduction of IDOFe³⁺ is sufficient to permit formation of this product. In the absence of H₂O₂ and O₂, V led to nearly quantitative reduction of the enzyme to IDOFe²⁺ (λ_max = 559 nm; isosbestic points, 461.5 nm and 523 nm) (Fig. 5). No other product of indole oxidation was able to reduce IDOFe³⁺, so we attribute the formation of IDOFe²⁺ observed during oxidation of indole by IDOFe³⁺ and H₂O₂ to reduction of IDOFe³⁺ by V after its formation.

Fig. 5.

Anaerobic reduction of IDOFe³⁺ by 3-oxoindole. IDOFe³⁺ (7 μM) was incubated with 3-acetoxyindole (70 μM) under N₂. Reaction was initiated by addition of esterase (2 units/mL) to generate 3-oxoindole. Spectra were recorded at 1-min intervals. (Inset) kinetics of indigo blue formation (608 nm) (●) and of enzyme reduction [ε_559nm = 5.23 mM^-1cm⁻¹ (IDOFe³⁺) and 15.03 mM^-1cm⁻¹ (IDOFe²⁺)] (○).

Discussion

Our results demonstrate that IDO has peroxygenase activity toward indole but not toward Trp. This reaction proceeds under aerobic and anaerobic conditions and is not inhibited by superoxide dismutase (SOD), d-mannitol, or lidocaine (SI Materials and Methods). Thus, IDO-catalyzed oxidation of indole in this reaction does not require O₂ or involve O₂^•−, ^•OH, or ¹O₂ outside the heme cavity. The consumption of indole and H₂O₂ proceeds at nearly stoichiometric proportions at moderate [H₂O₂], and the oxygen incorporated into the monooxygenated products originates from H₂O₂. In these experiments, the enzyme was initially present as IDOFe³⁺, and this state of the enzyme was regenerated when reaction was complete, thereby accounting for two oxidizing equivalents from H₂O₂, as expected for a peroxygenase. However, 3-oxoindole (V) formed initially can reduce IDOFe³⁺ to IDOFe²⁺ under specific conditions of [O₂] and [H₂O₂]. Thus, the characteristics of this IDO reaction closely resemble those of the “H₂O₂ shunt” reported for P450 (22–26) and lead us to propose that a reactive compound I intermediate (Fe⁴⁺=O, with a porphyrin or protein radical center) transfers the ferryl oxygen to indole to generate monooxygenated products.

Although oxidation of indole by IDO has not been reported previously, oxidation of indole by other heme enzymes [e.g., HRP (27), chloroperoxidases (28), P450s (29–31)] is well known; however, the mechanisms and products of these reactions differ. Nevertheless, catalysis by both oxygenases and peroxidases involves conversion of the ferric enzyme to the principal reactive intermediate compound I (32–35).

No direct evidence for the formation of IDO compound I is currently available, although evidence of a ferryl species with a protein-based radical has been reported for the related enzyme Trp 2,3-dioxygenase (36). Resonance Raman experiments indicating formation of IDO compound II after reaction of IDOFe³⁺ with H₂O₂ provide the only evidence for formation of a higher oxidation state by this enzyme (9, 11, 12). Oxygenation of the neutral indole by IDO compound II is unlikely, however, given that transfer of the ferryl oxygen would generate IDOFe²⁺ and lead to IDOFe³⁺-O₂^•−, which is unreactive with indole. Similarly, single electron oxidation of indole by IDO compound II followed by oxygen transfer to the ensuing indolyl radical by a second IDO compound II is unlikely in view of the near-stoichiometric consumption of indole and H₂O₂. Although we propose involvement of a compound I intermediate in the oxidation of indole by IDO, the location of the radical on the porphyrin or the protein is speculative. Because only compound II has been detected at present, dissipation of the putative radical and/or transfer of the oxygen from IDO compound I to indole must be prompt. Such tightly coupled transfer of oxygen to indole presumably involves formation of a transient compound I–indole complex in which the substrate can access ferryl oxygen. The inhibition of indole oxidation by l-Trp, which binds to IDO compound II without reacting (9), is consistent with formation of such a complex.

The transfer of oxygen to indole by IDO compound I (Fig. 6) to produce I and V could occur by hydroxylation or epoxidation, as known to occur for many reactions of P450 (25, 26). Hydroxylation of either C₂ or C₃ is presumably a consequence of the large active site of IDO, which would allow binding of indole so that either carbon atom can access the ferryl center. Once formed, V can react with O₂ to form VI, leading to indigo blue (15), or to form III (16). Two equivalents of IDOFe³⁺ alone can oxidize V to indigo blue, with the accumulation of IDOFe²⁺, presumably as the result of two one-electron oxidations. In the presence of H₂O₂, the ferryl/Fe³⁺ couple can promote rapid and extensive oxidation of V, again by two one-electron transfer reactions, consistent with the near absence of oxygenated products derived from anaerobic oxidation of V by H₂O₂ and IDOFe³⁺. In contrast, dioxygenated products III and IV can be derived by aerobic oxidation of I by IDOFe³⁺ in the presence of H₂O₂. The amount of ¹⁸O incorporated into IV by catalysis of H₂¹⁸O₂ oxidation of indole by IDO was consistent with near-stoichiometric incorporation of ¹⁸O at C₂ and partial (30%) incorporation of ¹⁸O at C₃. This isotopic distribution could result from the time-dependent exchange of ¹⁸O with H₂¹⁶O, but alternative sources for the second oxygen atom are possible.

Fig. 6.

Proposed reactions for the IDOFe³⁺ catalyzed of indole oxidation by H₂O₂.

Formation of II (an analogue of N–FK) is notable, because this compound is the only product that we have identified that has not been reported to result from P450-catalyzed oxidation of indole (29–31, 37–39). In contrast, oxidation of indoles (40) and some indoleamines (41) by HRP can lead to oxidative indole ring opening by reaction of indolyl radical with O₂ or O₂^•− (40), as reported for the photooxidation of l-Trp to N-FK (42). In principle, a peroxidatic mechanism could account for IDO oxidation of indole, as has been proposed for IDO oxidation of melatonin (8), but this possibility is ruled out for indole oxidation by incorporation of ¹⁸O from H₂¹⁸O₂. On the other hand, peroxidation could account for IDO-catalyzed ring opening of 3-methylindole and 2,3-dimethyl indole, neither of which incorporated ¹⁸O from H₂¹⁸O₂. Given that IDO lacks catalase activity (6), it is unlikely that sufficient ¹⁸O₂ forms by disproportionation of H₂¹⁸O₂ to react with a presumed indolyl radical and account for the substantial ¹⁸O content of the product. Instead, it is more likely that oxidation of indole to II and other indole products occurs along a common pathway initiated by the transfer of oxygen from IDO compound I.

Failure to observe II when I or V was used as substrate implies that neither of these compounds is an intermediate in formation of this product from indole. One possible intermediate is the highly labile indoline-2,3-epoxide (43–45), which decays to I- and V-type species (44, 45), consistent with the nearly identical incorporation of ¹⁸O into these two products. Recently, l-Trp epoxide has been proposed to accept an oxygen atom from IDO compound II during catalysis to regenerate IDOFe²⁺ (46, 47). A similar mechanism could convert the epoxide (before rearrangement) to II. The substoichiometric incorporation of ¹⁸O as the second atom of oxygen in II (and IV) is not consistent with this mechanism, however. One alternative mechanism involves nucleophilic attack on the epoxide by H₂O, which generally leads to a vicinal diol (48–50), in this case indoline-2,3-diol. This mechanism could permit incorporation of one atom of ¹⁶O from bulk solvent or ¹⁸O from H₂¹⁸O (resulting from H₂¹⁸O₂ after initial compound I formation). Indoline-2,3-diol may be unstable (51), but it could be oxidized by H₂O₂ and IDO-Fe³⁺ to form the stable products II and/or IV. The mechanism of this subsequent oxidation is speculative but may involve P450-type ferryl chemistry relevant to C-C bond cleavage of diols (26, 52) and alcohol oxidation (53, 54) (Fig. S4).

The remarkable selectivity that allows IDO to act on indole as a peroxygenase but on indoleamines exclusively as a dioxygenase reinforces the importance of the derivatized alkyl substituent in orientation of the substrate and possible stabilization of catalytic intermediates at the active site of this enzyme. As a result, electron transfer from indole to IDOFe³⁺-O₂^•− to produce a more highly oxidized form of IDO heme iron, as believed to occur with Trp (9, 11–13, 46, 55) does not occur. Similarly, no appreciable electron transfer from Trp bound near the heme iron to the presumed IDO compound I intermediate occurs to initiate substrate oxidation in the peroxygenase reaction. Without more detailed structural or spectroscopic information, it seems likely that the proposed ability of Trp to bind to more than one location at the active site of IDO is relevant to these observations (7, 56). Specifically, differences in the binding of Trp and indole to secondary sites at the active center of IDO may influence the binding orientations of both substrates in proximity to the heme iron and contribute significantly to the catalytic selectivity and the product distribution that we have observed.

Materials and Methods

Proteins and Reagents.

Bovine erythrocyte SOD, porcine liver esterase, glucose oxidase (Aspergillus niger), 30% H₂O₂, indole, 3-methylindole, 2-methylindole, 2,3-dimethylindole, 2-oxoindole, 3-methyl-2-oxoindole, 3-acetoxyindole, indole-2,3-dione (isatin), d-glucose, lidocaine, d-mannitol, and Trizma base were obtained from Sigma-Aldrich. Reagent concentrations were determined spectrophotometrically [H₂O₂, ε₂₄₀ = 39.4 M⁻¹cm⁻¹ (57); indole, ε₂₇₆ = 5,670 M⁻¹cm⁻¹ (58); 3-methylindole, ε₂₈₀ = 5,160 M⁻¹cm⁻¹ (58); 2-methylindole, ε₂₇₀ = 6,500 M⁻¹cm⁻¹ (58); 2,3-dimethylindole, ε₂₈₀ = 6,360 M⁻¹cm⁻¹ (58); 3-acetoxyindole, ε₂₇₈ = 5,500 M⁻¹cm⁻¹ (this study)]. Porcine liver esterase, glucose oxidase, and SOD were quantified from the activity reported by the manufacturer. IDO was expressed in Escherichia coli and purified (A₄₀₄/A₂₈₀ = 2.05) by a method culminating with elution over a Cibacron 3GA column (Sigma-Aldrich) (59). IDO concentration was determined spectrophotometrically (IDOFe³⁺ ε₄₀₆ = 175,000 M⁻¹cm⁻¹) (6).

Spectroscopy.

Electronic absorption spectra (20 °C) were recorded with a Cary 4000 or 6000i spectrophotometer. Fluorescence emission spectra (20 °C) were recorded with a Cary Eclipse spectrofluorimeter [slit widths, 20 (emission) and 1.5 nm (excitation)].

IDO-Catalyzed Oxidation of Indoles.

Typically, reactions (30 μL) were performed in Tris buffer (20 mM, pH 7.5, 20 °C), initiated by the addition of H₂O₂ [0.5–6 mM; natural abundance H₂O₂ or 90% (vol/vol) ¹⁸O-enriched H₂¹⁸O₂ in H₂¹⁶O; ICON Isotopes] to solutions containing IDO and indole (0.2 and 2 mM, respectively). Reactions were stopped after 60 s by addition of ice-cold methanol (50 μL). After centrifugation to remove inactivated enzyme, the supernatant fluid (20 μL) was analyzed immediately by HPLC. The lower aqueous solubilities of methylindoles required that they be dissolved in methanol (∼1–2%), which limited the concentrations of these compounds that could be assayed (1, mM for 3-methylindole, 1 mM for 2-methylindole, and 0.2 mM for 2,3-dimethylindole). Other derivatives assayed included 2-oxoindole (2 mM) and 3-oxoindole (1 mM). 3-Oxoindole was prepared from 3-acetoxyindole by anaerobic deesterification with porcine liver esterase (2 U/mL) immediately before use (60). Reactions were performed under aerobic conditions unless indicated otherwise. Anaerobic reactions were performed in a N₂-filled glove box (O₂ ≤1 ppm) using indole and protein solutions that had been purged with N₂ for ∼30 min. H₂O₂ solutions were deoxygenated by repeated freeze–pump–thaw cycles. In some cases, d-glucose (10–100 mM) and glucose oxidase (24–47 U/mL) were included to scavenge traces of O₂ (61). To monitor the progress of anaerobic reactions, reaction mixtures were prepared in Thunberg cuvettetes in a glove box. Anaerobically generated reaction products were collected in the glove box after the protein was precipitated with N₂-purged methanol, and the resulting reaction mixtures were clarified by centrifugation and analyzed by aerobic HPLC and ESI-MS (SI Materials and Methods and Figs. S5 and S6).

Kinetics of Indole Oxidation.

Solutions of IDO (1 μM) and indole (12.5–400 μM) in Tris buffer (20 mM; pH 7.5) were mixed with H₂O₂ (25 μM to 4 mM) (final volume, 300 μL) in a dual-path length masked cuvette (1 cm × 1 cm), and indole consumption was monitored (triplicate or greater) by fluorescence at 345 nm (λ_ex = 280 nm). The contribution of the primary product, 2-oxoindole, to fluorescence was negligible at this concentration. Nonlinearity of the fluorescence response was corrected with a standard curve ([indole] vs. λ_em = 345 nm). Initial rates were determined with the logarithmic approximation method (62), and kinetic parameters were obtained by nonlinear regression fitting of the data (Origin 8.0; OriginLab).