UV Resonance Raman Characterization of a Substrate Bound to Human Indoleamine 2,3-Dioxygenase 1

Abstract

Human indoleamine 2,3-dioxygenase 1 (IDO) is a heme enzyme that catalyzes the first reaction of the main metabolic pathway of L-tryptophan (Trp) to produce N-formylkynurenin. The reaction involves cleavage of the C₂=C₃ bond in the Trp indole ring and insertion of two atomic oxygens from the iron-bound O₂ into the indole 2 and 3 position. For establishment of the chemical mechanism of this unique enzymatic reaction, it is necessary to determine the conformation and electronic state of the substrate Trp bound to IDO. In this study, we measured the ultraviolet resonance Raman spectra of IDO in the presence of Trp to detect the vibrational modes of the substrate Trp. We compared the ultraviolet resonace Raman spectra of Trp in a ternary complex (Trp-bound cyanide enzyme) and a binary complex (Trp-bound reduced enzyme) of IDO with that of free Trp in solution and found that binding to IDO influences the conformation of Trp, resulting in similar changes in the two complexes, especially around the C₃-C_β bond. However, the presence of the diatomic ligand at the heme sixth coordination site in the ternary complex significantly alters the mobility and electronic structure of Trp, most likely resulting in the C₂=C₃ bond cleavage in the enzymatic reaction.

Significance

Indoleamine 2,3-dioxygenase 1 is a heme enzyme converting L-tryptophan (Trp) to N-formylkynurenin. In this study, we detected and compared the resonance Raman spectra of substrate Trp in the indoleamine 2,3-dioxygenase 1 CN⁻-bound ternary complex and binary complex. The substrate Trp in the ternary complex had a conformation similar to that in the binary complex. However, the existence of a diatomic ligand caused alteration in the electronic state, appearing as a difference in Raman spectrum. Especially, the difference found in the vibrational mode, including C₂, where oxygen binds and the C₂=C₃ bond cleaves during enzymatic reaction, was remarkable. Significant findings included detection of the clear difference in the conformational restriction and electronic structure in the substrate Trp between the binary and ternary complexes by using UVRR spectroscopy.

Introduction

Human indoleamine 2,3-dioxygenase 1 (IDO) is a heme-containing dioxygenase (1) that is found in various tissues of mammals (2) as an inducible enzyme (3). IDO, along with tryptophan 2,3-dioxygenase (TDO), catalyzes the direct incorporation of two oxygen atoms from molecular oxygen (O₂) to L-tryptophan (Trp), which is converted into N-formylkynurenine and metabolized through the kynurenine pathway. IDO constitutes a major target of cancer immunotherapy owing to its role in immune escape (4, 5, 6). Therefore, the establishment of its reaction mechanism based on structural information is necessary for not only basic enzymatic chemistry but also medical applications such as drug design (7). Traditionally, a dioxetane mechanism and Criegee rearrangement have been proposed as a possible reaction mechanism (8, 9, 10, 11). However, the first report of the crystal structure of IDO (12) revealed that it might not be possible for amino-acid residues in the heme pocket to abstract a proton from the N₁H of the substrate indole ring (the atomic numbering is shown in Fig. S1), although this step would be essential for the dioxetane mechanism and Criegee rearrangement. The density functional theory calculation (13) also indicated that neither mechanism was plausible owing to the presence of the high energy barrier.

Visible resonance Raman (RR) spectroscopy, which revealed the coordination geometries of several IDO intermediates based on the detection of Fe-O stretching modes during the catalytic reaction (14, 15, 16), considerably advanced research in the IDO reaction mechanism. Fig. 1 depicts the proposed reaction mechanism based on the unveiled heme coordination geometries of several intermediates that occur during the enzymatic reaction of IDO. In the reaction, oxygen (Fig. 1, b–f) directly binds to heme iron, whereas Trp as a substrate (Fig. 1 c) or intermediate products (Fig. 1, d–f) bind to a substrate-binding site in the distal heme pocket rather than directly binding to heme iron. In this figure, the heme coordination structures directly observed by visible RR are illustrated in red. For example, Fe-O₂ stretching was observed at 569 cm⁻¹ in Fig. 1 b (oxygenated intermediate) and Fe-O₂ and O-O stretching at 564 and 1138 cm⁻¹, respectively, in Fig. 1 c (Trp-bound oxygenated intermediate, hereafter the O₂-ternary complex). The detection of an Fe⁴⁺=O stretching RR band at 798 cm⁻¹ for Fig. 1 e (ferryl oxo intermediate) (14, 15) indicated the occurrence of a two-step oxygen atom transfer: one from Fe³⁺-O₂⁻ in Fig. 1 c and the other from Fe⁴⁺=O in Fig. 1 e. This is distinct from the reactions of heme-containing monoxygenases, in which only the proximal oxygen atom, forming Fe⁴⁺=O(por+ ⋅) or the corresponding oxidation state, is transferred to the substrate, whereas the terminal oxygen atom is converted to H₂O (9, 10). Accordingly, these findings provided an assumption that the substrate bound to the enzyme would play important roles in the unique IDO reaction. Consistent with this, recent computational studies of the reaction mechanism (17, 18, 19) suggested that the intermediate product bound to IDO as a substrate in Fig. 1 e (ferryl oxo intermediate) constitutes a Trp epoxide, which might facilitate the second oxygen transfer. However, it appeared experimentally difficult to detect the occurrence of such an intermediate product in the enzymatic reaction (the black colored illustration in Fig. 1). Consequently, it was necessary to obtain information regarding the conformation and electronic state of the substrate Trp bound to IDO, which would help to understand the chemical essence of the unique consecutive monooxygenation reactions mediated by IDO.

Figure 1.

Heme coordination geometries in some IDO intermediates (a–c, and e) determined by visible RR spectroscopy and proposed reaction mechanisms based thereon. Substrate Trp or intermediate product binds to a substrate-binding site in the distal heme pocket but not directly to the heme iron. States referenced in the main text are labeled. Illustrations colored in red indicate that the parts have already been observed directly by visible RR, whereas those shown in black indicate that the parts have not been observed. (d) and (f) in brackets indicate that these represent intermediates or transition states predicted according to the whole reaction. (g) and (h), labeled as “binary” and “ternary,” respectively, are the states examined in this study. NFK represents N-formylkynurenin.

Recently, the crystal structure of the Trp- and CN⁻-bound form of oxidized IDO, hereafter the “ternary complex” (Fig. 1 h; (20, 21)), was reported, which determined the conformation of Trp as a substrate at alkaline pH (pH 10). In this study, to explore additional details regarding the electronic structure, along with the Trp conformation in IDO at pH 8—a value relevant to the enzymatic reaction—we attempted to measure the vibrational spectra of the substrate Trp bound to IDO at the substrate-binding site using RR spectroscopy enhanced by ultraviolet (UV) excitation (UVRR), which constitutes a powerful tool for focusing aromatic compounds in proteins. We chose ternary and binary complexes of IDO (Fig. 1, h and g, respectively) as the UVRR samples, with the former being regarded as a model similar to the O₂-ternary (Fig. 1 c) complex as an intermediate appearing in the IDO reaction and the latter taken for comparison. Utilizing isotope double-difference techniques, we successfully obtained the vibrational spectra of the substrate Trp bound to IDO in the binary and ternary complexes. Comparison of these spectra with that of free Trp in solution was then applied to unveil effects of the surrounding protein portion and the iron diatomic ligand on Trp conformation and electronic structures.

Materials and Methods

Protein and chemicals

Recombinant IDO was produced using an Escherichia coli expression system and purified as explained elsewhere (12). Potassium cyanide and unlabeled Trp were purchased from Wako Pure Chemicals (Osaka, Japan). Labeled d₅-Trp (all carbons of indole deuterated) and 2-¹³C-Trp were purchased from Cambridge Isotope Laboratories (Tewksbury, MA).

Solution condition and sample preparation for spectroscopic measurements

Sample solution containing 100 μM IDO and 0.3 M NaClO₄ in 300 μL of 100 mM phosphate (pH 8.0) was transferred into an NMR sample tube of 9 inches in length and 5 mm outer diameter (535-PP-9SUP; Wilmad Lab Glass, Vineland, NJ). The Raman samples also contained 3 mM Trp (unlabeled Trp, d₅-Trp, or 2-¹³C-Trp) for the binary complex (Fig. 1 g) and 1 mM Trp (unlabeled Trp, d₅-Trp, or 2-¹³C-Trp) and 400 μM KCN for the ternary complex (Fig. 1 h). These concentrations were determined based on their dissociation constant as described in the Supporting Materials and Methods. We confirmed the lack of detectable difference in UVRR, visible absorption, and visible RR spectra upon the addition of 0.3 M NaClO₄ for both binary and ternary complexes, indicating a minimal and negligible effect of ClO₄⁻ ions on IDO structure in our experimental conditions. A stirring magnet bar (2 mm in length and 2 mm diameter) was also placed in the sample tube. The tube was sealed with a septum (Sleeve stopper septa 2.5 × 4.3 mm bottom; Sigma-Aldrich, St. Louis, MO), and the internal gas was replaced with N₂. For the samples in the reduced form, sodium dithionite solution was added to the final concentration of 3 mM under N₂ atmosphere. Samples of free Trp were prepared in the same manner as those of protein solution. The spectra of free Trp with and without dithionite were identical.

UVRR settings and measuring methods

The 229 nm emission from an Ar⁺ laser (300C FreD; Coherent) was used to excite Raman scattering. The excitation light at 0.5 mW was line-focused on the sample, and the back-scattered light along 135° from the excitation light was collected and focused on an entrance slit of the prepolychromator (LPM-200; Bunkou Keiki, Tokyo, Japan) and then led into the main single polychromator (SPEX 1000M; SPEX Industries, Edison, NJ) attached with a liquid-nitrogen-cooled CCD (400B/LN; Roper Scientific, Tucson, AZ). The NMR tube as a sample cell was spun at 240 rpm along with an internal stir bar that was held by a nearby magnet.

Because it was necessary to obtain double-difference spectra between binary or ternary complexes and free Trp for isotope-labeled Trp as well as unlabeled Trp, four samples, at least, needed be measured in one experiment to calculate the double-difference spectrum. The sample conditions, along with laser conditions for the four measurements, were carefully maintained to be constant. To eliminate the effects of longtime fluctuations of the laser, two spectra for which the difference was supposed to be calculated were measured in close vicinity of time. Upon accumulation of spectra, each of the four spectra were accumulated for 5 min on a single exposure, and the 5 min measurements were repeated sequentially to reach the total accumulation time of 60 min for individual species. During the sequential measurements, the samples waiting to be tested were retained in ice-cold water. Before and after the 5 min accumulation, visible absorption spectra were measured as described in the Supporting Materials and Methods to confirm the lack of perturbation of the sample. When alteration in the absorption spectrum was found, the UVRR spectra were discarded, and the measurements were stopped for the particular sample. The 60 min accumulation for each of the four samples is hereafter referred to as one measurement. After one measurement, the enzyme usually retained >90% of its catalytic activity, as described in the Supporting Materials and Methods.

UVRR spectra presented below represent the sum of three measurements, i.e., accumulation for 3 h for each sample unless otherwise indicated. For the calculation of a difference spectrum based on the Raman intensity of ClO₄⁻, which served as the internal intensity standard, we checked the linear correlation between the relative Raman intensity of Trp at the concentration of up to 4 mM against that of 0.3 M ClO₄⁻.

Isotope double-difference calculation method

The Raman spectrum of the IDO binary or ternary complex contained 1) Raman spectrum consequent to aromatic residues in IDO, 2) Raman spectrum owing to Trp bound to IDO as a substrate, and 3) Raman spectrum resulting from free Trp in solution. To extract the Raman spectrum owing to Trp bound to IDO as a substrate from the Raman spectrum of the IDO binary or ternary complex, an isotope double-difference spectral technique was utilized in this study. The addition of isotope-labeled Trp to the IDO sample solution yielded only labeled Trp bound to IDO as a substrate and free Trp in solution, rather than aromatic residues in IDO. All of the following difference calculations were performed based on the band intensity of ClO₄⁻ in each Raman spectrum.

Step 1

Calculation for isotope difference spectra of free Trp in solution. uW, dW, and cW were used to represent the UVRR spectra of free Trp in solution (without IDO) prepared with unlabeled, d₅-, and 2-¹³C-Trp (Fig. 2, A–C). Their difference spectra W_{u_d} and W_{u_c} (Fig. 2, B′ and C′) were defined as follows:

$$uW–dW=W_{u\_d},$$ (1)

$$uW–cW=W_{u\_c}.$$ (2)

Figure 2.

UVRR spectra of free Trp in solution with three types of isotope labels: (A) unlabeled Trp, (B) d₅-Trp, and (C) 2-¹³C-Trp along with isotope difference spectra (B′) = (A − B) and (C′) = (A − C). The most relevant vibrational modes between unlabeled and d₅-Trp are connected by a solid line revealing a large isotopic shift, whereas the most relevant vibrational modes between unlabeled and 2-¹³C-Trp have the same numbering with a small isotopic shift. The band at 934 cm⁻¹ due to the ClO₄⁻ ion serves as an internal intensity standard for calculation of the difference spectra. Raman intensities of the spectra presented in this figure are normalized by the ClO₄⁻ band intensity. Symbols uW, dW, W_{u_d}, cW, and W_{u_c} in the figure are quick references for the isotope double-difference calculation method provided in the main text.

Step 2

Calculation for isotope difference spectra of binary and ternary complexes. uB, dB, and cB were used to represent the UVRR spectra of binary complexes prepared with unlabeled Trp (Figs. 3 B and S2 A), d₅-Trp (Fig. S2 B), and 2-¹³C-Trp (Fig. S2 C). Their difference spectra B_{u_d} (Fig. S2 B′) and B_{u_c} (Fig. S2 C′) were defined as follows:

$$uB–dB=B_{u\_d},$$ (3)

$$uB–cB=B_{u\_c}.$$ (4)

Figure 3.

UVRR spectra of (A) reduced IDO, (B) binary complex, and (C) ternary complex prepared with unlabeled Trp. Spectra (B) and (C) are also referred as uB and uT in the isotope double-difference calculation method.

In turn, uT, dT, and cT were used to indicate the UVRR spectra of ternary complexes prepared with unlabeled Trp (Figs. 3 C and S3 A), d₅-Trp (Fig. S3 B), and 2-¹³C-Trp (Fig. S3 C). Their difference spectra T_{u_d} (Fig. S3 B′) and T_{u_c} (Fig. S3 C′) were defined as follows:

$$uT–dT=T_{u\_d},$$ (5)

$$uT–cT=T_{u\_c}.$$ (6)

As a result, 1) Raman spectra of aromatic residues in IDO, which were not isotope-labeled, were canceled; and 2) the Raman spectrum owing to Trp bound to IDO as a substrate and 3) the Raman spectrum resulting from free Trp in solution remained in the difference spectra.

Final step

Isotope double-difference spectrum: Extraction of the isotope difference spectrum of Trp as a substrate. Here, further differential calculations between resulting difference spectra from Eqs. 3, 4, 5, and 6 in step 2 and from Eqs. 1 and 2 in step 1 were performed. Isotope difference spectra of Trp as a substrate in binary complexes between unlabeled and d₅-Trp, SB_{u_d} (Figs. 4 A and S4 C), and unlabeled and 2-¹³C-Trp, SB_{u_c} (Fig. S5 A), were defined as follows:

$$B_{u\_d}–W_{u\_d}\left(2.9\mathrm{mM}\right)=S{B}_{u\_d},$$ (7)

$$B_{u\_c}–W_{u\_c}\left(2.9\mathrm{mM}\right)=S{B}_{u\_c}.$$ (8)

Figure 4.

Extracted isotope difference spectra between unlabeled Trp and d₅-Trp of Trp as substrate for (A) binary complex (Fig. S2B′ − Fig. 2B′ = Fig. S4A − Fig. S4B = Fig. S4C) and (B) ternary complex (Fig. S3B′ − Fig. 2B′ = Fig. S4D − Fig. S4E = Fig. S4F). The corresponding isotope difference spectrum of free Trp is also shown by (C), which is same as Fig. 2B′ but represented for comparison’s sake. Positive and negative peaks arise from the enzyme-bound unlabeled Trp and d₅-Trp, respectively. The lack of Wd18 (d₅-Trp) negative peak around 710 cm⁻¹ in the spectrum (A) is probably due to weakness and broadening of Wd18 of d₅-Trp. Spectra (A–C) are also referred as SB_{u_d}, ST_{u_d}, and W_{u_d} in the isotope double-difference calculation method provided in main text.

In comparison, the isotope difference spectra of Trp as a substrate in ternary complexes between unlabeled and d₅-Trp, ST_{u_d} (Figs. 4 B and S4 F), and unlabeled and 2-¹³C-Trp, ST_{u_c} (Fig. S5 B), were defined as follows:

$$T_{u\_d}–W_{u\_d}\left(0.9\mathrm{mM}\right)=S{T}_{u\_d},$$ (9)

$$T_{u\_c}–W_{u\_c}\left(0.9\mathrm{mM}\right)=S{T}_{u\_c},$$ (10)

where the spectral intensities of free Trp for binary and ternary complexes were adjusted to that of 2.9 mM free Trp for binary (Eqs. 7 and 8) and 0.9 mM for ternary (Eqs. 9 and 10) complexes by multiplying the obtained spectrum in step 1 by a factor because the concentrations of added free Trp differed. These calculations were guaranteed by the linear correlation between Raman intensity of free Trp in solution relative to that of 0.3 M ClO₄⁻ and the concentration of free Trp up to 4 mM.

Results

UVRR spectra of IDO in the absence and presence of Trp

Fig. 2 shows the UVRR spectra of free Trp excited at 229 nm along with the band assignments (22). The spectra of unlabeled Trp, d₅-Trp, and 2-¹³C-Trp are presented as Fig. 2, A–C, respectively, and their assignments are denoted as W, Wd, and Wc with subsequent mode numbers. These spectra were identical to those previously reported (23). In these spectra, the band from ClO₄⁻ ions observed at 934 cm⁻¹ served as an internal standard, the intensity of which was utilized for calculations of difference spectra in this experiment. The most relevant vibrational modes between unlabeled Trp and d₅-Trp (Fig. 2, A and B) are connected by solid lines, indicating that all bands in Fig. 2 B exhibit a large frequency shift upon deuteration of the indole ring in Trp. To clarify such shifts, their differences from the unlabeled Trp are also presented in the form of the difference spectra as Fig. 2 B′ (B′ = A − B in Fig. 2) and Fig. 2 C′ (C′ = A − C in Fig. 2). This part corresponds to step 1 in the double-difference calculation method described in the Materials and Methods. For example, the isotope shift of the W3 to Wd3 modes upon the H → D isotope substitution, 27 cm⁻¹, was sufficiently large to identify the original peak frequencies (Fig. 2, A and B) from the positive and negative peaks at 1552 and 1525 cm⁻¹ in the H–D difference spectrum (Fig. 2 B′). In contrast, the W7 band region in the H–D difference spectrum (Fig. 2 B′) is complicated. The lower-frequency counterpart (1342 cm⁻¹) of the W7 doublet band (Fig. 2 A) is located in a similar region to the Wd6 band at 1335 cm⁻¹ (Fig. 2 B), and its higher-frequency counterpart (1361 cm⁻¹; Fig. 2 A) is located in a similar region to the Wd5 band at 1382 cm⁻¹ (Fig. 2 B). Owing to such signal overlap, the H–D difference spectrum was not utilized to unambiguously identify the original band position, especially of the W7 doublet band.

In the ¹²C → ¹³C isotope label at the C₂ position, the observed spectrum (Fig. 2 C) and its vibrational modes are almost identical to those shown in Fig. 2 A except W3 and W18. The spectrum (Fig. 2 C′) reveals a clear and large differential pattern for W3 and a smaller one for W18. These observations confirmed that the W3 and W18 modes arise from the C₂=C₃ stretching and C₂ deformation characteristics, respectively. The isotope shift of W3 to Wc3 mode upon ¹²C → ¹³C substitution, 16 cm⁻¹, was too small to allow the frequency identification of the original peaks (Fig. 2, A and C) from the difference spectrum (Fig. 2 C′). Thus, we instead utilized the extracted H–D isotope difference spectrum for identification of the original peak frequencies.

Fig. 3 shows the UVRR spectra of IDO in the reduced (Fe²⁺) form (Fig. 3 A) and binary (Fig. 3 B) and ternary (Fig. 3 C) complexes, whose modeled structures are presented in Fig. 1, a, g, and h, respectively. Fig. 3 A represents the first report, to our knowledge, of a spectrum providing information regarding structures of the aromatic residues (Trp and Tyr) of IDO, denoted as W and Y, respectively, followed by their assigned mode numbers (22). In comparison, in addition to the signals from the six Trp residues in the IDO protein, the spectra shown in Fig. 3, B and C contain the Trp Raman bands from both Trp bound to the enzyme as a substrate and free Trp in solution.

The relative band intensity of ClO₄⁻ in the binary complex spectrum (Fig. 3 B) is weaker than that in the ternary complex spectrum (Fig. 3 C); the former sample was prepared with 3 mM Trp and the latter with 1 mM Trp, owing to the disparity in their dissociation constant. Because the band intensity of free Trp relative to the ClO₄⁻ band intensity has a linear correlation with the free Trp concentration, the contribution from free Trp in solution to the obtained Raman spectrum of the binary complex is considered to be larger than that in the ternary complex (see also Fig. S4). Because the concentration of IDO for the Raman measurement was 0.1 mM, contribution from the Trp bound to the enzyme in the binary and ternary complexes can be estimated at ∼0.1/3.6 and 0.1/1.6, respectively. The >twofold lower contribution of Trp at the low ratio might constitute a primary reason for the lower signal/noise ratio in the extracted isotope difference spectrum of substrate Trp in the binary complex compared to that in the ternary complex (Fig. 4, see below).

To investigate the conformation and electronic state of the substrate Trp bound to IDO in binary and ternary complexes, we extracted the vibrational signals of the substrate Trp by following the isotope double-difference calculation method. We calculated the isotope (H–D and ¹²C–¹³C) difference spectra for binary (Fig. S2) and ternary (Fig. S3) complexes, which corresponds to step 2 in the method (see Materials and Methods). In the calculated H–D and ¹²C–¹³C isotope difference spectra for the binary (Fig. S2, B′ and C′) and ternary (Fig. S3, B′ and C′) complexes, Tyr and therefore Trp Raman bands, owing to IDO amino-acid residues, were canceled as expected, whereas the Raman bands from free Trp in solution and those from Trp as a substrate were retained in the spectra. This suggests that the three different kinds of isotope-labeled Trp equally affect IDO structure. To highlight the signals from the substrate Trp bound to IDO, we subtracted the isotope difference spectrum of free Trp in solution (Figs. 2 B′ and S4, B and E) from the isotope difference spectra of the binary (Figs. S2 B′ and S4 A) and ternary (Fig. S3 B′ and S4 D) complexes. This constitutes the final step in the method (see Materials and Methods). In the resultant calculated spectra shown in Fig. 4, the isotope difference signals arising from the substrate Trp bound to IDO between unlabeled Trp and d₅-Trp are displayed for the (Fig. 4 A) binary and (Fig. 4 B) ternary complexes. These spectra are also shown in Fig. S4, in which the intensities of the difference spectra are normalized by ClO₄⁻ for reference. For comparison, the isotope difference spectrum of free Trp is also depicted as Fig. 4 C, which is the same spectrum illustrated in Fig. 2 B′. These three spectra in Fig. 4 are shown based on W16 band intensities for ease of comparison. The same calculation method was applied to obtain the ¹²C–¹³C isotope difference spectra of the substrate Trp bound to IDO in the binary and ternary complexes (Fig. S5), which were derived from the Raman spectra with unlabeled and 2-¹³C-Trp.

Vibrational modes of the substrate Trp bound to IDO

All the observed Raman bands in the spectra shown in Figs. 2 and 4 C are due to vibrational modes from free Trp in solution. When comparing this spectrum (Fig. 4 C) with those of the substrate Trp bound to IDO (Fig. 4, A and B), we found that several of these modes exhibit changes in position and intensity upon binding to IDO, whereas others do not. The changes should reflect the change of the conformation of Trp and its interaction with surroundings such as protein portion and the iron-bound ligand upon binding to IDO. A description of how the Trp vibrational modes were changed is as follows.

W3 and W18 modes

In the spectra comparison, it was found that the most notable changes were observed for the W3 and W18 modes, which arise from the C₂=C₃ stretching and indole ring breathing (stretching vibration of the whole indole ring as if it is breathing), including the C₂ deformation characteristics, respectively. Their frequencies were all reproducible within 1 cm⁻¹ in several independent measurements and well fitted with a Gaussian function (Fig. S6) within 1 cm⁻¹ error for the position (see below). With respect to the W3 mode for unlabeled Trp, the peak position was shifted from 1552 to 1545 or 1547 cm⁻¹ upon binary or ternary complex formation, respectively, as shown in Fig. 4 and Table 1. (Fig. S5 also shows the change in the Wc3 frequency from 1553 to 1548 or 1547 cm⁻¹.) Similar changes were also observed for the W18 band, the frequency of which was shifted from 759 to 766 or 767 cm⁻¹ upon complex formation. Therefore, these findings suggested that conformation and/or electronic interaction of Trp with its surroundings, especially around the C₂=C₃ bond, likely somewhat differed between the free and bound forms.

Table 1.

Assigned χ^2,1 Angle for each Trp in this Study along with Observed Frequencies of W3 and Wd3 and Plausible |χ^2,1|-Values Deduced from Empirical Equations

	Assignment of χ2,1/°	Observed Raman Shift/cm−1		Plausible |χ2,1| from Raman Shift/°
		W3	Wd3	W3	Wd3
Free	+98	1552	1525	22/98	23/97
Binary	−40	1545	1520	40/80	(33/87)a
Ternary	−34	1547	1521	34/86	31/89

^aBecause of the broad feature with lower signal/noise, Wd3 for binary complex gave lower reproducibility and thus is shown with parentheses.

To further clarify the conformational and/or electronic interaction changes related to the W3 and W18 bands, simulation fitting with a Gaussian function was performed, as depicted in Fig. S6 and summarized in Table S1. The simulation not only confirmed the frequency of the band positions but also quantified the band intensity and width (full width at half maximum) of these modes. The resultant band positions were the same as the readings of individual peaks. The simulation showed that the band widths were also altered by binding to IDO; the W3 band width became slightly broader from 18 to 21 cm⁻¹ upon binary complex formation, albeit unambiguously narrower to 9 cm⁻¹ upon ternary complex formation. The W18 band became narrower, from 18 to 13 cm⁻¹ upon binary complex formation and to 10 cm⁻¹ upon ternary complex formation (Table S1). Therefore, if inhomogeneous broadening is effective, these observations suggested that the substrate Trp in the binary complex exhibits mobility similar to that of free Trp in solution, whereas the mobility is rather restricted in the ternary complex.

In addition to the change of the band width, the band intensities of the W3 and W18 bands, relative to that of W16, were also changed; from free Trp to binary and ternary complex formation, the W3 changed from 1.2 to 2.0 and 0.5, whereas the W18 changed from 1.3 to 0.9 and 0.8, respectively. We chose W16 as a scale because the band was not overlapped with others, and its width and position were not altered by binding to IDO. The band intensity of Trp is affected by numerous factors such as the electric charge surrounding C_α, hydrophobicity around the indole ring, and the protonated or deprotonated states of its amino and carboxy groups. Thus, uneven relative intensity changes of the Trp band by binding to IDO in each complex indicates the occurrence of unique alterations in interaction with the surroundings for each complex.

W7 mode

The W7 vibrational mode normally appeared around 1360 and 1340 cm⁻¹ as a doublet, owing to the Fermi resonance between the W7 band corresponding to the N₁-C₈ stretching and to some combination bands of indole ring out-of-plane modes (22). In Fig. 4, the difference spectral pattern in the W7 region vary across the three spectra, i.e., the intensity and band positions of W7 and Wd6 in the three spectra differ. Despite such a complicated feature, the most obvious characteristic in this region is the clear peak at 1370 cm⁻¹ for the ternary complex (Fig. 4 B), indicating that the W7 higher-frequency counterpart shifted to ∼1370 cm⁻¹ via ternary complex formation. This band is not observed in the binary complex spectrum. A shoulder at 1370 cm⁻¹ is evident in the original UVRR spectrum of the ternary complex (Figs. 3 C and S3 A) but not in that of the reduced IDO (Fig. 3 A), binary complex (Fig. 3 B), or free Trp (Fig. 2 A).

W1, W16, and W17 modes

According to the analysis of the vibrational modes of Trp (22), the W1 mode predominantly corresponds to benzene C=C stretching with contribution from N₁C₈ stretching, the W16 mode corresponds to benzene ring breathing, and W17 corresponds to a mixed mode of benzene vibration and N₁H motion. In this manner, the W1, W16, and W17 vibrational modes at 1619, 1011, and 881 cm⁻¹, respectively, in Fig. 4 exhibit common aspects that mainly include vibration of the benzene ring. Although the band position of these modes did not change upon formation of the complexes, their relative intensity differed unevenly, especially the much stronger intensity of W1 in the binary complex (1.2) than that of the others in the ternary complex (0.1) and free Trp (0.2), as shown in the simulation fitting with the Gaussian function in Fig. S6 and Table S1. As was observed for W3 and W18, the different relative intensity of W1 reflects different interaction of the substrate Trp with its surroundings in the free, binary, and tertiary complexes.

Discussion

We determined the UVRR spectra of Trp bound to IDO as a substrate (Figs. 4 and S5). The extracted isotope difference spectra in Fig. 4, A and B are generally unique to each species and distinct from that of free Trp (Fig. 4 C) and likely reflect structural and environmental differences between free Trp and Trp as a substrate. The IDO protein portion and diatomic ligand appears to have several effects on Trp conformation and interaction.

Conformation of Trp in the bound state

The W3 vibrational mode, which involves the C₂=C₃ stretching characteristic, exhibited the most marked change in its frequency position upon binary and ternary complex formation. This mode is reported to be coupled with other modes (22) through a change of the C_α-C_β-C₃-C₂ torsional angle (χ^2,1; see Fig. 5 A) around the C_β-C₃ bond (24, 25), and eventually there is a strong correlation between the W3 frequency and χ^2,1 angle in the range of χ^2,1 = 0 ± 180°. Therefore, the W3 frequency was used to determine the χ^2,1 angle of Trp (24, 25, 26). According to the correlation, it would be possible to determine the χ^2,1 angle in free Trp and the substrate Trp bound to the binary and ternary complexes of IDO and obtain information regarding the conformation of the indole ring of Trp for each state.

Figure 5.

The definition of the C_α-C_β-C₃-C₂ torsional χ^2,1 angle with the viewpoint of Trp conformation focusing on χ^2,1 angle (A). Illustrations of obtained χ^2,1 angle in this study from the viewpoint in (A) are shown for free Trp in solution (B) and Trp as a substrate in IDO (C).

Using the reported equation and parameters (24, 25), the curve for unlabeled Trp was reproduced in Fig. 6 A for the range of |χ^2,1| = 0–120°, in which the majority of the reported Trp is found, with the observed frequencies for unlabeled Trp marked therein. When the reported empirical relationship was applied to our observed frequencies of the W3 mode, we obtained plausible values of |χ^2,1| from the curve, as shown in Fig. 6 and summarized in Table 1; 22 or 98° for free Trp (1552 cm⁻¹), 40 or 80° for the substrate Trp in the binary complex (1545 cm⁻¹), and 34 or 86° for the substrate Trp in the ternary complex (1547 cm⁻¹). The error for each Raman shift was less than ±1 cm⁻¹, and therefore the error for each angle was less than ±3°. A parallel correlation was also reported between Wd3 and |χ^2,1| as shown in Fig. 6 B (27). |χ^2,1| based on the Wd3 frequency was estimated as 23 or 97° for free Trp (1525 cm⁻¹), 33 or 87° for the substrate Trp in the binary complex (1520 cm⁻¹), and 31 or 89° for the substrate Trp in the ternary complex (1521 cm⁻¹). Despite the different correlation curves, the obtained values of |χ^2,1| are in good agreement between unlabeled Trp and d₅-Trp and can be considered based on the values observed for unlabeled Trp. As discussed in further detail below, we assigned the most plausible χ^2,1-values for each Trp, which are summarized in Table 1. The conformational difference focusing on the χ^2,1 angle is illustrated in Fig. 5.

Figure 6.

Relationship between the C_α-C_β-C₃-C₂ torsional angle |χ^2,1| and frequencies of W3 band: (A) for unlabeled Trp (W3), and the equation used is ν = 1542 + 6.7(cos 3 |χ^2,1| + 1)^1.2; (B) for d₅-Trp (Wd3), and the equation used is ν = 1514 + 7.5(cos 3 |χ^2,1| + 1)^1.2. The above equations are taken from (24) and (27), respectively. Observed Raman shifts and plausible |χ^2,1| are shown in brackets. The |χ^2,1|-values highlighted in bold with “x” were used for the final assignments.

Concerning the χ^2,1 of the free Trp, we adopted +98° (Fig. 5 B) from the predicted |χ^2,1|-values in Table 1 because of good consistency with the ¹H NMR results of Trp in solution (28). In comparison, for Trp bound to IDO as a substrate, we proposed that the χ^2,1 angle of the binary complex is −40° and that of the ternary complex is −34° (Fig. 5 C) on the basis of structural information obtained by several other techniques. For example, T. Kamachi at the Fukuoka Institute of Technology in Japan carried out a docking simulation of Trp using the MacroModel program and IDO crystal data (Protein Data Bank [PDB]: 2D0T) and showed that χ^2,1 = −41° was most stable under the fully flexible conditions of protein and ligand and that a negative χ^2,1 angle is appropriate for the substrate L-Trp (personal communication). The x-ray data of TDO and IDO, which reveal a similar arrangement of amino-acid residues in the heme pocket (20, 29), may help to determine the χ^2,1 of the substrate Trp in the binary and ternary complexes of IDO based on our UVRR data. χ^2,1 = −40 and −44° were revealed for each dimer structure in the crystal of the binary complex of bacterial TDO (29), whereas χ^2,1 angles at −34 and −38° were reported for the O₂-ternary complex of human TDO (30). The crystal structure of the TDO binary complex was obtained at pH 5.6 (29), whereas that of the TDO O₂-ternary complex was obtained at pH 6.3 (30). The crystal structures of the IDO ternary complex, which were prepared at pH 10, also indicated that the χ^2,1 angles were −52 and −60° for two monomers in the asymmetric unit (20) and that the angle varied between −44 and −58° upon mutation or addition of an effector (20, 21).

By comparing these χ^2,1-values between the crystallographic and UVRR data, we found that the χ^2,1-values fall in the same range, although a slight difference was observed upon detailed comparison. The gap in the observed values between these two methods might be due to differences in measurement conditions such as pH value or buffer type, along with differences in crystalline or solution states. For example, because IDO exhibits the highest reactivity at pH 6.5, the bound Trp at pH 10 (the crystallization condition of IDO) is presumed to be flexible even in the crystal. This might contribute to the larger χ^2,1 angle in the crystal structure. Alternatively, it should be noted that both UVRR for IDO and crystallographic data for TDO revealed that the χ^2,1-value decreased from the binary to the ternary complexes. This is unambiguously caused by the binding of the external ligand (CN⁻) to the heme iron sixth coordination site of IDO, as discussed in the following section.

Interaction of Trp with its surroundings in IDO

In addition to the conformational differences of Trp revealed by the torsional angle χ^2,1, the band widths of the W3 and W18 modes, including C₂ displacement, differ among Trp in the three different states (see Table S1), implying that mobility around the Trp χ^2,1 angle differs among several states: i.e., in the binary complex, it might be similar to that of free Trp, whereas it becomes smaller upon ternary complex formation. These differences appearing with regard to Trp conformation, as well as its mobility upon complex formation, likely resulted from the interaction of Trp with IDO. Possibilities include 1) polar interaction including hydrogen bonding network and 2) hydrophobic interaction including cation-π and π-π interactions of Trp; these are considered on the basis of the Raman spectra obtained in this study as follows.

Interaction of carboxy and amino groups of Trp

It has been reported that the protonation state, i.e., the protonated or deprotonated state, of both amino and carboxy groups of Trp affects Raman intensities of many UVRR bands such as W1, W3, and W18, along with the frequency of W18 (31). Because alteration in the protonation state of these two functional groups was interpreted as alteration in an electronic charge around C_α, the band intensity and/or frequency change can be explained in terms of an alteration in the polar interaction with surroundings, including the hydrogen bonding state of these two functional groups. For example, the W18 band frequency differs by ∼10 cm⁻¹ depending on the protonation state of the carboxy group, whereas the W1 band intensity is affected by the protonation state of the amino group but not by that of the carboxy group. These criteria should be utilized during elucidation of the interaction of the Trp carboxy and amino groups with their surroundings.

The carboxy group of Trp with pK value of 2.46 (32) is deprotonated in our experimental condition (pH 8). Nevertheless, we found that the W18 band at 759 cm⁻¹ in the free Trp exhibited a higher-frequency shift to ∼766 cm⁻¹ in both the binary and ternary complexes. This might reflect a difference of the interaction of the Trp carboxy group between the free form in solution and the bound form to IDO. Consistent with this, the crystal structures of both IDO (20) and TDO (29, 30) revealed that the carboxy group of the substrate Trp forms an ion pair with Arg231 (or equivalent Arg144 in TDO) and hydrogen bonding with Thr379 (or equivalent Thr342 in TDO). It is also noted that these interactions are unaltered regardless of the existence of an iron sixth ligand in the TDO crystal. Nevertheless, although Arg231 has been reported as a critical residue for substrate binding in IDO, Arg144 in TDO is less critical (33).

Notably, the W1 band intensity was much stronger in the binary complex than those in the ternary complex (see Figs. 4 and S6). This likely reflected a different hydrogen bonding state of the Trp amino group in these two complexes. The amino group in Trp with a pK value of 9.41 (32) most likely exists in the protonated form in our experimental condition, whereas the ternary complex (Fig. 1 g) has a diatomic ligand (CN⁻ in this case) at the heme iron sixth coordination site that is not present in the binary complex. According to the visible RR studies of CN⁻- and CO-bound IDO, Trp binding as a substrate to IDO strengthens the Fe-CN⁻ and Fe-CO bonds and weakens the C-O bond concomitant with alteration in the heme side chain (34, 35). The band widths of νFe-CO and νC-O become much narrower in the CO-ternary complex as compared to those in CO-bound IDO without substrate in visible RR (35). These results indicated that an interaction is present between the diatomic ligand and Trp as a substrate in the ternary complex, and that the interaction likely contributes to the mobility of the iron-coordinated ligand. An electron nuclear double resonance and electron paramagnetic resonance study (36) also suggested the presence of the interaction between the diatomic ligand of Fe and the substrate Trp in the ternary complex; moreover, the crystal structure of the ternary complex showed that not only CN⁻ and heme propionate but also Thr 379 are within a distance capable of forming a hydrogen bond with the amino group of the substrate Trp (20). Together, these observations clearly indicate that the amino group in Trp as a substrate in the ternary complex strongly interacts with CN⁻, heme propionate, and Thr through hydrogen bonding and/or electrostatic interaction. This is likely attributable to the mobility of the ligand along with the bound Trp (see below).

Interaction of the indole N₁H

The frequency of W17 is known to be sensitive to hydrogen bonding of the N₁H group (37). However, the lack of change in Raman shift at 881 cm⁻¹ upon complex formation with IDO reasonably demonstrated that the N₁H group of Trp as a substrate does not have a strong hydrogen bond with amino-acid residues in the protein. This is consistent with the crystal structure of the ternary complex (20), in which no amino-acid residues interact with the N₁H group. Although the distance between CN⁻ and N₁H is as close as 3 Å, their angle may not be suitable for hydrogen bonding. The electron nuclear double resonance (36) study supported this suggestion. In the crystal structure of the ternary complex, H₂O (W660) is present at a short distance and likely forms a hydrogen bond with the indole N₁H (20). However, the lack of change in the W17 Raman shift upon formation of the ternary complex indicated that the environment around the indole N₁H is likely unaltered from those of the free Trp in aqueous solution and the binary complex.

Hydrophobic interaction

The crystallographic study of IDO in the ternary complex has shown that some hydrophobic residues such as Tyr126, Val130, Phe163, and Leu234 exist around the Trp indole ring (20). The nearest residue is Phe163, which points to the pyrrole ring of Trp in the edge-to-face manner with a short distance (<3.4 Å without hydrogen). Tyr126, Val130, and Leu234 are located at 3.5, 3.6, and 3.8 Å away from the indole ring without hydrogen, respectively. These observations suggested that a hydrophobic interaction of the indole ring of the substrate Trp occurs in IDO, most probably both for the binary and ternary complexes. In this regard, several mutation studies have been reported for the hydrophobic residues in the IDO distal heme pocket (12, 33, 38). Single mutations of Phe to Ala in the distal pocket do not affect the K_d values of Trp in reduced IDO or CN⁻-IDO, whereas an effect of these mutations including F163A on enzymatic activity is observed, although it is dependent on experimental conditions (12, 33, 38). In the UVRR spectra, it has been generally known that the hydrophobicity of the indole ring strongly affects the Raman intensity for W3, albeit to a lesser extent for W7 and W16 (31). In our UVRR of the substrate Trp in IDO, we found that the W3 band intensity largely differed between the binary and ternary complexes, as shown in Fig. 4, implying a difference in the strength of the hydrophobic interaction of the indole ring with its surroundings in these two complexes.

With respect to the strength of the hydrophobic interaction of Trp in proteins, some UVRR studies of other proteins have shown that the Raman shift of the W7 mode may serve as a good marker (39, 40, 41, 42, 43). For example, the spectrum of bacteriorhodopsin yields W7 as a triplet, and its highest frequency component, derived from Trp182, appears at 1370 cm⁻¹ (40). Higher-frequency shifts of the W7 higher-frequency counterparts were also observed for Trp53 at 1369 cm⁻¹ in the direct oxygen-sensor protein from E. coli (41) and for Trp354 at 1370 cm⁻¹ in the Yersinia protein tyrosine phosphatase (42). These higher-frequency shifts in W7 are attributed to the steric repulsion consequent to the very strong hydrophobic interaction between Trp and its surroundings: Trp182 and the retinal methyl group (40); Trp53 and Val34, Leu68, Leu101, and the heme 2-vinyl group (41); and Trp354, Val360 and Arg409 (39), respectively. However, another interpretation has been proposed for the last case, in which Trp354 interacts with Arg409 through cation-π (43) to result in the higher-frequency shift of W7. A similar explanation has also been applied for Trp38 (43) in the filamentous virus Pf3, which exhibits a cation-π interaction between Trp38 and Arg39 or Lys40.

Notably, the higher-frequency counterpart of the W7 doublet in this study was found at 1370 cm⁻¹ in the IDO ternary complex. Together with the alteration of the relative intensity of the W3 modes in the binary and ternary complexes, our UVRR results indicated that the hydrophobic interaction between the indole ring of the substrate Trp with its surroundings is present in both complexes but is stronger and/or specific, such as a π-π or cation-π interaction, in the ternary compared to the binary complex. If a specific interaction with Phe163 appears as an edge-to-face interaction at the position nearest to Trp in the crystal structure (20), the altered electronic structure of the substrate in the ternary complex may explain the significant reduction in reactivity in the F163A mutant (33). Thus, the association of the diatomic ligand (CN⁻ in this case) might strengthen the hydrophobic interaction of the Trp indole ring, most likely through interaction with the Trp amino group, as stated above. As a result, mobility of the indole ring, especially in the χ^2,1 rotation, is limited in the ternary complex.

The interaction of the substrate Trp with IDO and its resultant conformation and electronic structure, which were elucidated by our present UVRR study, can be summarized as follows. Primary interactions of Trp with the protein portion of IDO include both hydrophobic interaction of its indole ring with Tyr126, Val130, Phe163, and Leu234 and the electrostatic or hydrogen-bonding interaction of its carboxy group (-COO⁻) with Arg231 and Thr379 along with amino (-NH₃⁺) and propionate groups. These interactions determine the C_α-C_β-C₃-C₂ torsional angle of the substrate Trp as being different from that of the free Trp in solution but comparable in the two complexes, thereby causing its C₂=C₃ bond, i.e., the insertion site of two atomic oxygens, to be close to the heme active center. In addition, the conformation of the substrate is consistent with the superior reaction of IDO with L-Trp, compared to that with D-Trp, at physiological concentrations (44). Conversely, the interaction might not be sufficiently strong to fine-tune the configuration of the substrate Trp in IDO, because its mobility in the binary complex is comparable to that of free Trp in solution. However, the binding of the sixth ligand (CN⁻) to the heme iron in the ternary complex dramatically reduces the mobility of Trp through the electrostatic interaction with the Trp amino group (-NH₃⁺). The interaction might strengthen and/or specialize the hydrophobic interaction of the Trp indole ring. Eventually, owing to the presence of the heme sixth ligand, the affinity of Trp likely becomes higher for the ternary than for the binary complex. In the physiological IDO dioxygenation reaction, the O₂ binding to the heme sixth coordination site may fix and activate the substrate Trp, as reflected by the unique alteration only in the W3 and W18 vibration modes in which C₂ displacement is most involved. The O₂ ligand may also be activated by the bound Trp with a unique electronic structure. The observed UVRR spectra of the substrate Trp experimentally confirmed that fine-tuning by its surroundings, including the heme, O₂, and protein, causes unique electronic and structural alteration suitable for the sequential IDO reactions.

Author Contributions

The manuscript was written through contributions of all authors.

Editor: Wendy Shaw.