How is the distal pocket of a heme protein optimized for binding of tryptophan?

Authors


Correspondence

E. L. Raven, Department of Chemistry, University of Leicester, Leicester LE1 7RH, UK

Fax: +44 0116 252 2789

Tel: +44 0116 229 7047

E-mail: emma.raven@le.ac.uk

Abstract

Indoleamine 2,3-dioxygenase and tryptophan 2,3-dioxygenase catalyze the O2-dependent oxidation of l-tryptophan to N-formylkynurenine. Both are heme-containing enzymes, with a proximal histidine ligand, as found in the globins and peroxidases. From the structural information available so far, the distal heme pockets of these enzymes can contain a histidine residue (in tryptophan 2,3-dioxygenases), an arginine residue and numerous hydrophobic residues that line the pocket. We have examined the functional role of each of these residues in both human indoleamine 2,3-dioxygenase and human tryptophan 2,3-dioxygenase. We found that the distal histidine does not play an essential catalytic role, although substrate binding can be affected by removing the distal arginine and reducing the hydrophobic nature of the binding pocket. We collate the information obtained in the present study with that reported in the available literature to draw comparisons across the family and to provide a more coherent picture of how the heme pocket is optimized for tryptophan binding.

Abbreviations
hIDO

human indoleamine 2,3-dioxygenase

hTDO

human tryptophan 2,3-dioxygenase

IDO

indoleamine 2,3-dioxygenase

rTDO

Ralstonia metallidurans TDO

TDO

tryptophan 2,3-dioxygenase

xTDO

Xanthomonas campestris TDO

Introduction

The first and rate-limiting step in the kynurenine pathway is the O2-dependent oxidation of l-tryptophan to N-formylkynurenine (Scheme 1). The enzymes that catalyze this reaction are referred to in the early literature as ‘tryptophan pyrrolase’ [1-3] (and even as ‘tryptophan peroxidase-oxidase’) [4]; only later did the nomenclature evolve more systematically to describe either one of the two types of enzyme that catalyze this same reaction: tryptophan 2,3-dioxygenase (TDO) or indoleamine 2,3-dioxygenase (IDO). Although both enzymes catalyze the same reaction, there are a number of differences between them: TDO is typically tetrameric and, in mammals, it is located primarily in the liver, whereas IDO is monomeric and located ubiquitously around the body, apart from in the liver. The substrate specificities are also different, as reflected in the nomenclature, with the IDOs having generally a broader substrate specificity for numerous indole-derived substrates compared to the more substrate-specific TDOs.

Scheme 1.

The reaction catalyzed by indoleamine 2,3-dioxygenase and tryptophan 2,3-dioxygenase. In 1-Me-Trp, the proton on N1 is replaced with a Me group; in tryptamine, the carboxylate group of the side chain is replaced with a proton.

A considerable volume of spectroscopic and functional studies was carried out on both IDO and TDO from the 1970s to the 1990s [5], after which time interest in these enzymes waned. The more recent realization that tryptophan oxidation is implicated in a wide range of disease states means that IDO and TDO find themselves once again in the limelight, and they are currently the subject of a great deal of interest [6, 7]. An important breakthrough was the structure for recombinant human IDO (hIDO) [8], which was crystallized with the inhibitor 4-phenylimidazole bound to the heme iron. The active site structure not only confirmed some of the earlier predictions from spectroscopic work, but also contained some unexpected revelations. Predictably perhaps, considering the nature of the substrate, the IDO active site contains a large number of hydrophobic aromatic residues, including in the distal pocket (Fig. 1A). Most unexpectedly, no histidine residue was found in the distal pocket, which was in direct contradiction to the predictions obtained from low-temperature spectroscopic work [9-11]. Indeed, the entire IDO active site is almost completely devoid of polar residues: Ser167 is the only polar residue (Fig. 1A), although two mutagenesis studies suggest that this is not essential for catalysis [8, 12]. The later crystal structure of a bacterial TDO from Xanthomonas campestris (xTDO) [13] to a large extent not only confirmed the generally hydrophobic nature of the dioxygenase active site (Fig. 1B), but also included information on the interactions that tie the substrate to the enzyme in the substrate-bound complex.

Figure 1.

Comparison of the active sites of (A) hIDO (2DOT) [8], (B) X. campestris TDO (2NW8) [13] and (C) R. metallidurans TDO (2NOX) [14]. The heme is shown in the same orientation in all cases. For the xTDO structure, the substrate (yellow) is also shown bound to the enzyme, with hydrogen bonds indicated by dotted lines.

A structure has also been published for Ralstonia metallidurans TDO (rTDO) [14], which has an active site structure similar to that of the hIDO and xTDO enzymes (Fig. 1C). There is as yet no full analysis of the individual contributions of various active site residues to the overall process of tryptophan oxidation. This is in contrast to other heme proteins (e.g. the globins and the peroxidases) [15, 16], for which very extensive studies have been carried out with the aim of dissecting the contributions of individual residues. In this sense, although all of the IDO and TDO enzymes that have been studied so far have active sites that are broadly recognizable as heme enzymes, and in some ways are structurally analogous to other heme proteins (such as the globins and the heme peroxidases), we still do not understand how the active site pocket is optimized for its specialized role in the oxidation of exogenous tryptophan bound in the active site (a reaction that, so far as is known, lies only within the gift of IDO and TDO). To begin to address this, we have examined the functional consequences of individual mutations in two human dioxygenases: hIDO and human tryptophan 2,3-dioxygenase (hTDO). The results obtained reveal some of the subtle influences that are imposed upon the molecule by different amino acids and, by comparison with known bacterial enzymes, we use the information to build a more comprehensive picture of the individual roles of key residues on the catalytic process.

Results

Comparison of active sites

The R. metallidurans active site structure [14] is similar to that of hIDO [8] and xTDO [13] (Fig. 1). Because the active site structures of these enzymes appear to be similar, mutational analysis of hIDO and hTDO, as reported in the present study, is presumed to be representative of the family as a whole. Active site residues targeted for mutagenesis in hIDO, based on the published structure, are shown in Table 1. For hTDO, no crystal structure is available. Hence, a model of hTDO, created using hidden Markov modelling and sequence homology, was calculated using the hhpred server [17] and the template structures were found to correspond to Protein Data Bank entries 2NOX (rTDO), 2NW8 (xTDO) and 2DOT (hIDO) (which are the same structures shown in Fig. 1). Superposition of these published structures and the model for hTDO suggest the spatial correspondence of the active site amino acid residues, and provide a rationalization for the mutations in hTDO shown in Table 1.

Table 1. Comparison of hIDO, hTDO, rTDO and xTDO active sites (see also Fig. 1), including the variants discussed in the present study. Pairs of residues considered to align are presented in the same row (e.g. S167 and H76 in hTDO and hIDO, respectively). Data on the H55A, H55S, S167A and S167H variants have been reported previously [12, 22]. The Y130F, F68A and R134A variants have been reported previously [14]
hIDOhTDOrTDOxTDO
ResidueVariantResidueVariantResidueResidue
S167S167A, S167HH76H76A, H76SH72H55
F226F226A, F226YF140F140A, F140YY130Y113
F163F163AF72F72AF68F51
R231R231KR144R144A, R144KR134R117
F164F164A   I52
F227F227A  S131M114

Ligand binding

Wavelength maxima for the various ferric and ferrous derivatives of the hIDO and hTDO variants are presented in Table S1. They are broadly similar to those for the corresponding wild-type proteins and are not discussed in further detail.

Equilibrium binding constants for binding l-Trp (KD,Trp) are presented in Table 2. For comparison, binding constants for binding of cyanide (KD,CN) are also presented and do not deviate markedly from wild-type values.

Table 2. Equilibrium binding constants (KD, μm) for the binding of l-Trp and cyanide to ferric hIDO, hTDO and variants (pH 8.0, 25.0 °C). Pairs of residues considered to align are presented in the same row (e.g. S167 and H76 in hIDO and hTDO, respectively)
 hIDO hTDO
KD,Trp (Fe3+)KD,CN (Fe3+)KD,Trp (Fe3+)KD,CN (Fe3+)
  1. a From Chauhan et al. [12]; the effects of these changes have been reported previously. Spectroscopic changes associated with the binding of l-Trp to the ferrous enzyme were generally not considered sufficiently large to allow an accurate KD value to be measured. The KD,Trp (Fe3+) values for these hIDO variants are broadly in line with those previously reported for hIDO from a previous study [8]: F226A (0.8 mm), F163A (1.08 mm) and F227A (0.65 mm). The values for the wild-type proteins between the two studies also correspond (0.32 mm) [8].

Wild-type290 ± 63.5 ± 0.7Wild-type160 ± 2031 ± 1
S167A7900 ± 8001.9 ± 0.2H76A230 ± 1011 ± 1
S167H6900 ± 1100590 ± 30H76S270 ± 3048 ± 3
F226A2200 ± 3001.5 ± 0.1F140A540 ± 6065 ± 7
F226Y540 ± 502.5 ± 0.3F140Y560 ± 6051 ± 3
F163A450 ± 301.2 ± 0.2F72A220 ± 2024 ± 2
R231K2600 ± 2002.3 ± 0.4R144K520 ± 6015 ± 1
   R144A1900 ± 20065 ± 3
F164A1400 ± 3002.4 ± 0.4   
F227A640 ± 202.0 ± 0.2   

hIDO

For the hIDO variants, most values of KD,Trp for the ferric protein were broadly similar to those for the wild-type, and the most substantial changes are confined to the F226A and R231K variants (KD,Trp increased by approximately 10-fold), which suggests a role for both of these residues in l-Trp binding. This is explored further in the steady-state analyses below.

hTDO

These effects for hIDO are largely duplicated in hTDO because none of the mutations had large effects on either l-Trp or cyanide binding (Table 2), with the exception of the R144A mutation, which resulted in an approximately 10-fold increase in KD,Trp.

Steady-state oxidation of l-Trp

Steady-state parameters for l-Trp oxidation are reported in Table 3; representative data sets are shown in Fig. 2 for hIDO and the F163A variant.

Table 3. Steady-state kinetic data for oxidation of l-Trp by hIDO, hTDO and variants (50 mm Tris-HCl, pH 8.0, 25.0 °C)
 hIDO hTDO
kcat (s−1)KMm)kcat (s−1)KMm)
  1. a From Chauhan et al. [12]. b Measureable activity was not observed under the conditions measured. The F72A variant was particularly unstable and denatured under normal steady-state assay conditions (pH 8.0); at pH 9, increases at A321 could be observed, although reproducibility at each l-Trp concentration was very poor and a full steady-state profile could not be extracted. Instead, we measured Vmax at a single concentration of substrate (2 mm) to extract an apparent kcat for both F72A and hTDO at pH 9.0 and found that the F72A variant had < 3% of the activity of hTDO [apparent kcat(hTDO) = 1.9 s−1; apparent kcat(F72A) = 0.06 s−1], which is comparable to the corresponding F163A variant in hIDO (2% of wild-type) activity.

WT1.4 ± 0.17.0 ± 0.8WT1.4 ± 0.01170 ± 20
S167A1.6 ± 0.121 ± 2H76A0.30 ± 0.01480 ± 50
S167H0.0060 ± 3 × 10−526 ± 1H76S0.10 ± 0.01600 ± 70
F226A0.39 ± 0.01940 ± 70F140A0.040 ± 0.001920 ± 90
F226Y6.0 ± 0.225 ± 2F140Y0.80 ± 0.02460 ± 40
F163A0.03 ± 0.00147 ± 5F72A 
R231K0.31 ± 0.013300 ± 400R144K0.16 ± 0.01830 ± 60
   R144A 
F164A0.74 ± 0.01210 ± 10   
F227A0.43 ± 0.019.0 ± 0.4   
Figure 2.

Steady-state oxidation of l-Trp by (A) hIDO and (B) the F163A variant of hIDO. Solid lines show a fit of the data to the Michaelis–Menten equation (conditions: 50 mm Tris/HCl, pH 8.0, 25.0 °C).

hIDO

For hIDO, the effects on l-Trp binding in the ferric enzyme (KD,Trp, as above) are also seen in the steady-state (which reports on l-Trp binding to the ferrous enzyme), although the effects are more pronounced because the F226A variant (an approximately 130-fold increase in KM) and the R231K variant (an approximately 470-fold increase) show severely decreased substrate binding affinities (note, however, that the F226Y variant retains its ability to bind substrate; Tables 2 and 3). More modest effects on substrate binding are observed for the F163A and F164A variants (approximately 10-fold and 30-fold increases in KM, respectively). Regardless of the effects on KM, values of kcat for most variants are largely unchanged compared to the wild-type protein, with the exception of the F163A variant where kcat is considerably reduced (approximately 50-fold slower). Overall, we interpret these effects as indicating that, although binding affinity is lowered in some cases, the rate-limiting steady-state turnover (i.e. the conversion of Trp to N-formylkynurenine) is less seriously affected by the mutations and remains close to wild-type levels in most cases.

hTDO

These observations for hIDO are broadly mirrored in hTDO, although the effects are less pronounced. By comparison to hIDO, both the F140A (assumed to be equivalent to F226A in hIDO) and the R144K (assumed to be equivalent to R231K) variants show similarly but more marginally (approximately five-fold) weakened affinity for l-Trp binding; the F72A (equivalent to F163 in hIDO) and R144A variants were found to be completely inactive and the data for R144A suggest that this may arise from a weakened affinity of the ferrous enzyme for substrate under steady-state conditions (we were unable to measure a KM for the F72A variant). The H76A variant (equivalent to S167 in hIDO and H55 in xTDO) has activity that is essentially unchanged from the wild-type enzyme (Table 2). The H76S variant shows a more marked effect on kcat (an approximately 10-fold decrease) but neither variant shows a substantial effect on KM. We interpret this as indicating that His76 does not play an essential role. This is discussed in more detail below.

Redox potentiometry

We have noted previously [9] that reduction potentials in the presence and absence of substrate are correlated with the binding affinity of the ferric and ferrous forms, which means that the measured reduction potentials report directly on the relative binding affinity of the substrate. In hTDO, the reduction potential does not shift upwards on the binding of substrate [18], which differentiates it from both hIDO and xTDO (both of which show upwards shifts in potential on binding of l-Trp) [9, 13]. The reduction potential data for the hTDO variants are in accordance with these conclusions because the measured Fe3+/Fe2+ reduction potentials for the hTDO variants in the presence and absence of substrate (Table 4; with a representative data set shown in Fig. 3) shows only minor increases in reduction potential (with the exception of H76A which shows a larger difference). We have interpreted this information [18] as indicating that hIDO needs to specifically favour binding to the reduced form, whereas hTDO does not discriminate in this way, possibly reflecting a requirement imposed upon TDO in the cellular environment (oxidizing versus reduced). Indeed, rather than exhibiting preferential binding of substrate to the reduced enzyme (as does hIDO), the binding of l-Trp to hTDO is typically much weaker (Table 2) and was previously suggested [19] to account for the fact that substrate inhibition is not observed in hTDO. The reduction potential data for the variants support these suggestions.

Table 4. Fe2+/Fe3+ reduction potentials (mV) obtained for hTDO and variants in the absence (−Trp) and presence (+Trp) of l-Trp. Reaction conditions are described in the Experimental procedures. Reduction potentials for hIDO variants have been reported previously [19]
 –Trp+Trp
hTDO−92 ± 3−76 ± 3
H76A−52 ± 2+2 ± 3
H76S−119 ± 4−113 ± 3
F140A−129 ± 5−116 ± 3
F140Y−156 ± 2−129 ± 4
F72A−140 ± 1−139 ± 1
R144K−143 ± 3−116 ± 3
R144A−133 ± 3−112 ± 3
Figure 3.

Redox potentiometry of the F72A variant of hTDO showing the Nernst plot in the absence of l-Trp (solid circles) and the presence of 3 mm l-Trp (open circles). Reaction conditions: 100 mm potassium phosphate (pH 7.0) at 25.0 °C.

Discussion

In cases where structural information is available for IDO or TDO enzymes (Fig. 1), the active sites reveal a handful of residues that can potentially regulate substrate binding and catalysis. The data available in the literature on the effects of active site mutations on tryptophan dioxygenase activity are neither extensive nor conclusive. The first study [20] looked at the role of His346, which is the proximal ligand in hIDO. Along with their landmark structure, Sugimoto et al. [8] provided very preliminary activity data for variants at a number of positions (C129A, F163A, S167A, F226A, F227A, R231A and S263A), although only KD values for substrate binding were reported. Two studies have looked at the role of the distal histidine [21, 22], although they differ in their conclusions. Additionally, the roles of F68, R134 and Y130 (Fig. 1C) have been examined briefly in rTDO [14], and the single T342A variant in hTDO has also been studied [23]. However, taken together, these investigations have not yet led to a clear view on the role of individual residues in the active site.

In the present study, we examined the likely roles of these active site residues in hIDO and hTDO and the main effects observed are summarized below.

The role of the distal histidine

The question of whether there is a role for an active site histidine in the mechanism has been muddled in the literature: early data supported its involvement [24], although the subsequent ground swell of opinion does not. The only experimental evidence against the proposal so far is that 1-Me-Trp (Scheme 1) is a substrate for hIDO, albeit a slow one [25]. Assuming that Me-Trp and Trp react by the same mechanism (which is not conclusively established), we have interpreted [25] this observation as indicating that base-catalyzed abstraction of the indole proton by histidine cannot be involved in the mechanism (because it is not possible to remove a Me group via the same mechanism). Recent computational data [26-29] and electron-nuclear double resonance spectroscopy [30] all support our suggestion.

The data reported in the present study help to build a more extensive and persuasive body of experimental evidence backing up our original proposals [25]. The data for hTDO show only minor effects on both kcat (approximately five- and 10-fold decreases) and KM (< 4-fold increases) for the H76 variants. Previously published data on xTDO (H55A, H55S) [22] are in agreement with these findings (<10-fold changes in kcat and negligible changes in KM on the removal of H55). Our interpretation of all of the data is that histidine is not involved in base-catalyzed abstraction and this would be consistent with our suggestions for the mechanism [25], which do not require abstraction of the indole proton. The fact that there is no histidine in hIDO appears to be in agreement with the general conclusions outlined above.

The distal arginine

The structure of xTDO in complex with substrate shows ionic interactions between the carboxylate group on the substrate and Arg117 (Fig. 1B). This residue is also present in hIDO (Arg231) and rTDO (Arg134) and, from our model, presumably also present in hTDO (Arg144).

Replacement of the active site Arg with Lys in both hIDO and hTDO has fairly minor (< 10-fold reduction) effects on kcat (Table 3), although substantially different consequences on substrate binding in the two enzymes. In hIDO, KM increases by approximately 500-fold in R231K and these changes are mirrored in the KD for Trp binding to the ferric form (an increase of approximately 10-fold in R231K). In hTDO, KM increases only by approximately five-fold in R144K, and this is mirrored in the KD for Trp binding to the ferric form; the R144A variant is found to be inactive, although the binding data for the ferric enzyme indicate that substrate binding is again affected by the mutation [KD,Trp (Fe3+) is increased by approximately 10-fold]. The corresponding variant in R. metallidurans TDO (R134A) is also reported as inactive, although no detailed kinetic analyses have been provided [14].

Our summary of all of the data is that the interaction with Arg affects substrate binding in all of the enzymes examined so far; it is critical for substrate binding in hIDO but is less critical in hTDO. For hIDO at least, there is only one (critical) hydrogen-bonding interaction (to Arg) that fits with this interpretation because the residues equivalent to Y113 and His55 in xTDO (which provide the additional stabilization) are not present (replaced with Phe and Ser, respectively; Table 1). There are consistent reports in the literature that the binding of Trp is sensitive to substitution of the carboxylate group on the substrate because tryptamine (Scheme 1) is reported as being inactive in three different TDOs [13, 14, 18] and in hIDO [18]. These previous observations for the TDOs are thus in agreement with the conclusions that we draw regarding the role of the distal Arg. For IDO, the picture is less clear cut. Early work [19] reports very low (< 0.5%) activity for rabbit IDO against tryptamine, although we have not been able to reproduce this activity for the hIDO isolated from the expression vector reported in the present study or with the same hIDO in another expression vector with a cleavable His-tag. In our hands, we found hIDO to be inactive towards tryptamine, which is in accordance with the conclusions that we draw from the data on TDO above. A word of caution may be required. These early reports [19] have been widely cited [5, 8, 13] as providing evidence that IDO turns over substrates other than the native substrate (l-Trp); for tryptamine at least, this may need to be reassessed.

Hydrophobic residues

Hydrophobic residues are in abundance in the active site (Fig. 1). The most useful information on their role comes from mutagenesis data on the hIDO variants.

Phe163

Phe163 is critical for activity in hIDO, as indicated by a dramatic decrease in kcat, and the effect of this residue is duplicated in hTDO because replacement of the equivalent residue (Phe72) inactivates the enzyme. Because the equivalent residue in R. metallidurans TDO (F68A) is also inactive (although no kinetic parameters were reported) [14], we conclude that Phe163 in hIDO and its equivalent residues in hTDO and R. metallidurans TDO are key determinants of activity. This is also supported by steady-state data for the F51A variant of xTDO (kcat,F51A approximately 0.05 × kcat,WT, data not shown).

Phe227

Although Phe227 forms part of the substrate binding pocket in hIDO, removal of Phe227 has a very marginal effect (Table 3). In our hands, the F227A variant is catalytically competent, and we conclude that this residue has a non-essential role, presumably because it is too far from the substrate binding site (Fig. 1A) to affect the hydrophobic interactions to the substrate.

Phe164 and Phe226

Reducing the hydrophobic nature of the distal pocket would be expected to affect binding of the largely hydrophobic substrate. The data for hIDO support this: the largest effects are seen for F226A (approximately 130-fold decreases in KM), with similar but smaller effects for F164A (an approximately 30-fold decrease). This is consistent with a role for these residues in providing π-π stacking interactions with l-Trp. Consistent with these conclusions, a conservative substitution at Phe226 (F226Y) is tolerated and has no substantial effect (Table 3). Neither F226A, nor F164A show a substantial drop in kcat and we conclude that the turnover activity of these enzymes is not affected by the mutation.

However, when the analysis is expanded more widely and other proteins are considered, it is found that the effects of Phe226 are not, overall, duplicated across other TDOs and an inconsistent pattern emerges. The F140Y substitution in hTDO is tolerated but the F140A substitution is not, and the corresponding Y130F variant of R. metallidurans TDO shows a modest increase in kcat/KM. Further data are needed before clarification of the complex role of hydrophobic residues can be fully understood.

Concluding remarks

In summary, the conclusions based on the data obtained in the present study, as well as those from previous studies, are: (a) the distal histidine, where present, plays a non-essential role; (b) the active site arginine has a role to play in substrate binding but is more influential in some cases (hIDO) than in others (hTDO); and (c) hydrophobic interactions can affect substrate binding, with the most substantial effects being observed in hIDO Phe226 and Phe164. Clearly, further mutagenesis work on other proteins will be required before a full picture emerges.

Experimental procedures

Materials

l-ascorbate, bovine liver catalase, DNase I, d-glucose, glucose oxidase, methylene blue, xanthine, xanthine oxidase, all redox potential dyes, l-tryptophan and substrate analogues were purchased from Sigma-Aldrich Co. (St Louis, MO, USA). Escherichia coli strains BL21 (DE3) and Rosetta (DE3) pLysS were obtained from Novagen (Madison, WI, USA). Hemin was obtained from Fluka (Buchs, Switzerland).

Mutagenesis and protein purification

Site-directed mutagenesis was performed in accordance with the Quickchange protocol (Stratagene Ltd, Cambridge, UK). All variants were expressed and purified in accordance with previously published procedures for the corresponding enzyme [12, 18]. Absorption coefficients for all variants were determined using the pyridine hemochromagen method [31].

Steady-state kinetic assays

Activities were measured spectrophotometrically by monitoring the formation of N-formylkynurenine at 321 nm. Steady-state kinetic measurements were carried out using either a Perkin Elmer Lambda 35 UV-visible spectrophotometer (Perkin Elmer, Boston, MA, USA) or a Cary 50-Probe UV-visible spectrophotometer (Varian Inc., Palo Alto, CA, USA). Reactions were performed at 25.0 °C in 50 mm Tris-HCl buffer (pH 8.0). Assays contained 10 μm methylene blue, 100 μg of catalase, 20 mm l-ascorbate and the appropriate amount of enzyme. The reaction was initiated by the addition of l-Trp and initial rates were calculated from the increase in A321321 = 3750 m−1·cm−1) [32]. Apparent KM and kcat values were determined by varying the concentration of each substrate and fitting the data to the Michaelis–Menten equation.

Ligand-bound derivatives

All ligand binding data were measured in 50 mm Tris-HCl buffer (pH 8.0) at 25.0 °C. Ligand bound derivatives (cyanide, l-Trp) were obtained by the addition (typically 2–10 μL) of a concentrated stock solution to the enzyme. Ferrous proteins were generated by stoichiometric titration of the ferric enzyme with sodium dithionite. Equilibrium binding constants, KD, were determined in accordance with previously reported procedures [33].

Redox potentiometry

Redox potentiometry was carried out in accordance with previously reported procedures [18]. Reduction potentials (Fe3+/Fe2+) were determined at 25.0 °C by the reduction of the protein with a dye of known potential [34]. The assay solution contained potassium phosphate buffer (0.1 m, pH 7.0), glucose (5 mm), xanthine (16 mm), xanthine oxidase (5 μm), glucose oxidase (50 μg·mL−1), catalase (5 μg·mL−1) (Sigma-Aldrich Co.) and enzyme (2 μm). Heme reduction potentials were determined by fitting the data to a Nernst equation for a single-electron process [34] using origin software (Microcal, Inc., Northampton, MA, USA).

Acknowledgements

The present study was supported by The Wellcome Trust (project grant 083636 to E.R. and equipment grant WT087777MA to E.R.). The PCR for TDO variants was carried out by X. Yang (University of Leicester, UK) and the peptide mass fingerprinting was carried out by A. Bottrill and S. Ibrahim (University of Leicester, UK).

Ancillary