Polar Interactions between Substrate and Flavin Control Iodotyrosine Deiodinase Function

Abstract

Flavin cofactors offer a wide range of chemical mechanisms to support a great diversity in catalytic function. As a corollary, such diversity necessitates careful control within each flavoprotein to limit its function to an appropriate subset of possible reactions and substrates. This task falls to the protein environment surrounding the flavin in most enzymes. For iodotyrosine deiodinase that catalyzes a reductive dehalogenation of halotyrosines, substrates can dictate the chemistry available to the flavin. Their ability to stabilize the necessary one-electron reduced semiquinone form of flavin strictly depends on a direct coordination between the flavin and α-ammonium and carboxylate groups of its substrates. While perturbations to the carboxylate group do not significantly affect binding to the resting oxidized form of the deiodinase, dehalogenation (k_cat/K_m) is suppressed by over 2000-fold. Lack of the α-ammonium group abolishes detectable binding and dehalogenation. Substitution of the ammonium group with a hydroxyl group does not restore measurable binding but does support dehalogenation with an efficiency greater than those of the carboxylate derivatives. Consistent with these observations, the flavin semiquinone does not accumulate during redox titration in the presence of inert substrate analogues lacking either the α-ammonium or carboxylate groups. As a complement, a nitroreductase activity based on hydride transfer is revealed for the appropriate substrates with perturbations to their zwitterion.

Graphical Abstract

INTRODUCTION

Flavin-dependent catalysis is diverse and widespread and encompasses both oxidative and reductive transformations including, but certainly not limited to, hydrogenation, hydroxylation, halogenation, nitroreduction, and dehalogenation.^1–3 The unusual breadth of flavin reactivity can be attributed to its ability to promote both single-electron and two-electron (hydride) transfers. The specific function of each enzyme is programmed by the environment surrounding the flavin to alternatively activate and suppress competing modes of reaction. Sterics, electrostatics, solvation, hydrogen bonding, and proton transfer in proximity to flavin are typically governed by the protein backbone and its side chains.^4,5 To activate single-electron transfers, hydrogen bond donation to the flavin N5 appears most critical (Figure 1),^6–8 and polar interactions to its N3^9,10 and carbonyl oxygens (O², O⁴)^11–13 typically contribute in subordinate roles.

Figure 1.

Polar contacts surround the flavin of IYD in the presence of a zwitterionic substrate or ligand.

The nature of the hydrogen bonding partners further dictates the type of chemistry catalyzed by flavoproteins. For example, the NfsA and NfsB subfamilies within the nitroreductase superfamily offer hydrogen bonding to the flavin N5 through their amide backbone to promote hydride transfer during reduction of nitroaromatics.^14,15 BluB and IYD subfamilies within the same superfamily offer hydrogen bonding to the N5 position from side chains of Ser and Thr to induce single-electron chemistry necessary for oxygen activation and dehalogenation, respectively.^16,17 Replacement of the Thr with Ala in iodotyrosine deiodinase (IYD) destabilizes the intermediate one-electron reduced flavin semiquinone (FMN_sq) and relieves suppression of l-3-nitrotyrosine (O₂N-Tyr) reduction presumably by hydride transfer (Scheme 1).¹⁸ Accordingly, the hydrogen bond donor to the flavin N5 was considered the key mechanistic determinant for IYD. At least for human IYD, this hydrogen bond is controlled by substrate binding and the subsequent closure of an active site lid (PDB 4TTB and 4TTC) (Figure 1).¹⁷ In a bacterial homologue, an equivalent lid closes around l-3-iodotyrosine (I-Tyr) to promote deiodination (PDB 5KO8, see the Supporting Information).¹⁹ 2-Iodophenol also adopted an equivalent geometry of stacking over the flavin as observed by crystallography (PDB 5KRD).¹⁹ However, the lid remained too dynamic in the presence of 2-iodophenol since no electron density associated with this region was observed. In accordance with the trend of one- versus two-electron transfer for this system, 2-iodophenol was dehalogenated very inefficiently while 2-nitrophenol was easily reduced.²⁰

Scheme 1.

Products of Single-Electron and Hydride Transfers Catalyzed by IYD^a

^a(A) Dehalogenation results from two successive one-electron transfers. (B) Nitroreduction results from hydride transfer and typically generates a hydroxylamine product after two turnovers. The oxidized, one-electron reduced, and two-electron reduced flavins are indicated by FMN_ox, FMN_sq, and FMN_hq, respectively.

Our initial emphasis on the hydrogen bonding to flavin N5 was supported from a preponderance of literature reports and a general emphasis on the role of proteins to control cofactor activity. Initial studies on IYD were consistent with this dogma. However, the relatively small decrease in dehalogenation efficiency (15-fold) was surprising when the N5 hydrogen bond was lost upon the Thr to Ala substitution.¹⁸ For context, dehalogenation was suppressed by ~16,000-fold when the native flavin was replaced with its 5-deaza derivative that is incapable of single-electron transfer.²¹ Similarly, the importance of the substrate-induced closure of the active site lid as first suggested by crystallography requires reevaluation based on recent NMR experiments that suggest both tyrosine and phenol derivatives induce similar changes to the active site lid and bind in similar environments.²² Furthermore, a new lid sequence was successfully engineered to close around 2-iodophenol, but only a modest gain in dehalogenation was achieved with this variant.²³ Together, these data suggest that other features of IYD may be most important for stabilizing the FMN_sq intermediate and promoting dehalogenation. The conserved Arg that interacts with the flavin N1 and C2 carbonyl (O²) had previously been deemed nonessential,²⁴ and thus attention next turned to the contributions of the substrate as described below (Figure 1). A series of zwitterion derivatives of I-Tyr, L-3-fluorotyrosine (F-Tyr), and O₂N-Tyr have been prepared and characterized with Haliscomenobacter hydrossis IYD to demonstrate the dominance of the substrate in controlling the chemistry of catalysis.

MATERIALS AND METHODS

General Materials.

3-Fluoro-l-tyrosine (F-Tyr) and (S)-3-(4-hydroxyphenyl)lactic acid (HPLA) were obtained from AstaTech, Inc. (Bristol, Pennsylvania). d-Tyrosine (d-Tyr), dl-tyrosine (dl-Tyr), di-tert-butyl dicarbonate (Boc₂O), and xanthine were obtained from Alfa Aesar (Ward Hill, Massachusetts). Formic acid (88%), N-α-benzoyl-dl-arginine-4-nitroanalide hydrochloride (dl-BAPNA), and trifluoroacetic acid were obtained from Acros Organics (Morris Plains, New Jersey). l-Tyrosinamide, tyramine, trypsin from bovine pancreas, sodium hydrosulfite, 3-(4-hydroxyphenyl)propionic acid, 1-chloromethyl-4-fluoro-1,4-diazoniabicyclo[2.2.2]octane bis(tetrafluoroborate) (SelectFluor), flavin mononucleotide (FMN), N-iodosuccinimide, ortho-phenylenediamine, sodium pyrophosphate tetrabasic decahydrate, xanthine oxidase, methyl viologen, 3-nitro-l-tyrosine, peroxidase from horseradish, l-amino acid oxidase (LAAO) from Crotalus adamanteus, and d-amino acid oxidase (DAAO) from porcine kidney were obtained from Sigma-Aldrich (Madison, Wisconsin). Sodium acetate, ammonium hydroxide (28%), and sodium nitrite were obtained from Thermo Fisher Scientific (Waltham, Massachusetts). Nitric acid (68%, ACS grade) was obtained from VWR Scientific (Radnor, PA). SilicaFlash P60 silica was used for all column chromatography (SiliCycle Inc.).

General Methods.

NMR spectra were recorded on Bruker Avance 300 or 400 MHz spectrometers. Chemical shifts are reported in parts per million (ppm), and coupling constants (J) are reported in hertz (Hz). The residual ¹H and ¹³C from solvent were used as references. ¹⁹F signals were referenced from an internal standard of trifluoroacetic acid. All UV–vis measurements were performed on an Agilent 8453 spectrophotometer. Fluorescence measurements were performed with a FluoroMax-4 spectrofluorometer (Horiba Scientific). Mass analysis was performed on a VG Analytical VG-70S magnetic sector mass spectrometer with electron impact (EI) ionization.

Expression and Mutation of Iodotyrosine Deiodinase.

IYD from Haliscomenobacter hydrossis was expressed with a C-terminal His₆-tag, purified by Ni²⁺ affinity chromatography, and stored as described previously.¹⁹ The E91K/K116E variant was generated by two rounds of site-directed mutagenesis first using a forward primer 5′-GCT TTC AAA TTT CTC TTT TTC AGC AGC TTG TCG G-3′ and a reverse primer 3′-AAA GAG AAA TTT GAA AGC TAT AAT GGC CGC ATG TCC-5′ for the E91K mutation and subsequently using the forward primer 5′-AAA TGG TTC ATG CCA ATC GGT TCC AAA AGG-3′ and reverse primer 3′-TGG CAT GAA CCA TTT TTG GAA ATT GCT CCG TAT CTT ATT G-5′ for the K116E mutation. Mutations were confirmed by DNA sequencing, and the desired plasmids were transformed into Escherichia coli Rosetta 2(DE3) cells for expression and purification following standard protocols.¹⁹

Catalytic Deiodination.

Dehalogenation was initiated by adding sodium dithionite (5%) in sodium bicarbonate (5%, 100 μL) to a solution of IYD (final 100–250 nM) and the indicated concentration of substrate in 100 mM potassium phosphate pH 7.4 (900 μL). After 1–3 h (25 °C), the reaction was quenched by addition of 88% formic acid (50 μL) and products were quantified by their A₂₈₀ during separation by reverse-phase HPLC using a Varian Microsorb MV 300–5 (C18 250 × 4.6 mm) column with a JASCO MW-1510 PDA.²⁵ HPLC solvent A was 0.44% aqueous formic acid, and solvent B was 0.44% formic acid in acetonitrile. The solvent program began with 95% solvent A (0−10 min) and then increased linearly to 40% solvent A (10−25 min) and then to 5% solvent A (30−41 min) and finally 100% solvent A (41−55 min) at 1 mL/min. Quantification of the deiodinated product was performed as described previously by comparing its integrated signal (A₂₈₀) with an internal standard (meta-cresol) that had been calibrated with each substrate.²⁶

Affinity of Ligands to IYD Containing FMN_ox.

Ligands were titrated into a solution of IYD (3 μM, 2.5 mL) in 110 mM potassium phosphate pH 7.4 that was equilibrated for 20 min at 25 °C with gentle stirring. Fluorescence was measured 2 min after each addition of ligand to allow for equilibration. Fluorescence emission at 516 nm was recorded using an excitation of 450 nm, and K_d values were calculated by plotting relative fluorescence (ΔF/F₀) with respect to total enzyme ([E]_T) and total ligand ([L]) using eq 1 (Origin 6.0).^19,27

$$\frac{F}{F_0}=1+\frac{\Delta F}{F_0}\left[\frac{\left(K_{\text{d}}+{\left[\text{E}\right]}_{\text{T}}+\left[\text{L}\right]\right)-\sqrt{{\left(K_{\text{d}}+{\left[\text{E}\right]}_{\text{T}}+\left[\text{L}\right]\right)}^2-4{\left[\text{E}\right]}_{\text{T}}\left[\text{L}\right]}}{2{\left[\text{E}\right]}_{\text{T}}}\right]$$ (1)

Redox Titration of IYD in the Presence of Substrate Analogues.

Xanthine (900 μM), methyl viologen (2 μM), and substrate (concentration at 10× its K_d) in 110 mM potassium phosphate pH 7.4 were added to a sealed quartz cuvette at 25 °C, as previously described.^17,28,29 The solution was bubbled with argon for 20 min to remove molecular oxygen, and then IYD (20 μM) was added. The solution was bubbled with argon for an additional 2 min followed by a flushing of the cuvette head space with argon for an additional 20 min. Reaction was initiated by xanthine oxidase (40 μg/mL), and spectra were recorded every 1 min for 3 h. Absorbance versus wavelength was exported with a background subtraction from absorbance from 800 to 900 nm and plotted with Origin 6.0. Titrations for measuring reduction potentials also included a redox dye (20 μM). Suitable dyes were selected based on their reduction potentials within approximately 30 mV of that measured for IYD;³⁰ anthraquinone-2,6-disulfonate (AQDS, E_m = −188 mV) and anthraquinone-2-sulfonate (AQS, E_m = −225 mV) were used for IYD in the presence and absence of substrate analogues, respectively.³¹ Reduction of IYD and dyes was measured at wavelengths corresponding to the isosbestic point of the other chromophore. The known reduction potential of the dye along with the relative proportion of oxidized/reduced dye, and IYD was used to calculate the reduction potential of IYD using eq 2, as described previously.^29,31

$$\text{log}\left(\frac{D_{\text{ox}}}{D_{\text{red}}}\right)=\frac{n_{\text{D}}F}{n_{\text{E}}F}\text{log}\left(\frac{E_{\text{ox}}}{E_{\text{red}}}\right)+\frac{\left(E_{\text{m}}\left(E\right)-E_{\text{m}}\left(D\right)\right)n_{\text{D}}F}{2.303RT}$$ (2)

D = dye, n = number of electrons, F = Faraday constant, E = enzyme, E_m = midpoint potential, R = ideal gas constant, and T = temperature. Reduction potentials are calculated from the average of two individual titrations.

Nitroreductase Activity of IYD.

A solution of potassium phosphate (110 mM, pH 7.4) was degassed in a quartz cuvette by bubbling with argon for 20 min at 25 °C, as previously described.¹⁸ IYD (15–18 μM) was added, and the solution (900 μL final) was bubbled with argon for an additional 2 min; then, the headspace of the cuvette was purged with argon for 20 min. Enzyme was reduced by titration with sodium dithionite (0.25%, 1 μL aliquots) to generate reduced IYD•FMN_hq, as monitored by A₄₅₀. Air-saturated buffer (20 μL) was added to ensure that a small fraction of the enzyme would reoxidize as expected when no excess dithionite was present. Aliquots of a substrate solution (4 μM) degassed with argon in a separate cuvette for 20 min were transferred to the enzyme-containing cuvette, and spectra were recorded every 30 s for 60 min until full reoxidation to FMN_ox was observed by A₄₅₀. Spectra were corrected for a background signal recorded from 800 to 900 nm.

Limited Proteolysis of IYD in the Presence of Substrate Analogues.

IYD (5 μg) and its active site ligand (at concentrations of 10× its K_d) were incubated in 110 mM potassium phosphate pH 7.4 (25 μL final) at 25 °C for 5 min as previously described.²³ Reaction was initiated by addition of trypsin (1:50 w:w), and aliquots (5 μL) were taken after the indicated times, quenched with 2× SDS loading dye (5 μL), and placed in a 90 °C heat block for 5 min. The aliquots were then analyzed with 15% SDS-PAGE (40 min, 200 V). Coomassie brilliant blue staining solution (100 mL) was mixed with the gel, microwaved for 45 s, and placed on a shaker for 15 min. The gel was destained after addition of an aqueous solution of 10% glacial acetic acid and 30% methanol (100 mL) followed by microwaving for 45 s and shaking for 12 h. The gel was scanned with an HP Scanjet G4050, and signals were quantified with ImageQuant Software. The signal corresponding to a control sample of IYD was normalized to 5 μg, providing the relative quantity of IYD at each time. The disappearance of the IYD was plotted vs time and fit to a first-order decay with Origin 6.0 to calculate the half-life. Each digestion was performed in duplicate.

Potential Inhibition of Trypsin by Ligands of IYD.

A solution of trypsin (8 μM) and ligand (at a concentration of 10× its K_d for IYD) in 110 mM potassium phosphate pH 7.4 (65 μL) was incubated at 25 °C for 5 min, as previously described.²³ Reaction was initiated by addition of DL-BAPNA (A₃₁₅ = 0.45).³² Formation of 4-nitroaniline, the hydrolysis product of DL-BAPNA, was quantified by monitoring its absorbance at 405 nm (ε = 9920 M⁻¹cm⁻¹).³² Spectra were recorded every 15 s for 5 min. The rate of 4-nitroaniline formation over the first 2 min of reaction was plotted vs time, and linear regression was performed in Origin 6.0. These analyses were repeated independently twice.

Synthesis of Substrate Analogues.

Detailed syntheses of substrate analogues are described in the Supporting Information. Briefly, iodination of tyrosine analogues typically was performed with iodine in ammonium hydroxide, but (S)-3-(3-iodo-4-hydroxyphenyl)lactic acid (I-HPLA) was prepared with N-iodosuccinimide.^33,34 Fluorination of N-Boc-protected tyrosine analogues utilized SelectFluor.³⁵ Nitration of tyrosine analogues typically used nitric acid, but the lactic acid derivative, (S)-3-(4-hydroxy-3-nitrophenyl)lactic acid (O₂N-HPLA), was generated by diazotization and hydrolysis of O₂N-Tyr with sodium nitrite and sulfuric acid.^36–38

RESULTS AND DISCUSSION

Effects of Zwitterion Modification on Substrate Binding and Turnover.

Most commonly, the role of polar interactions in controlling the chemistry of flavin is tested by site-directed mutagenesis since the overwhelming number of these interactions involves proteins. For IYD, significant contributions are proposed to involve the zwitterion of its substrates (Figure 1). Accordingly, a series of halotyrosine derivatives were prepared to systematically delete or modify the α-ammonium and carboxylate groups without altering the native protein (Figure 2). The most active substrates for human IYD are 3-iodo-, bromo-, and chlorotyrosines, and all generate tyrosine after dehalogenation.^39–41 This profile is maintained in other IYDs from fruit flies and thermophilic bacteria.^25,42 Because tyrosine binds relatively weakly to IYD relative to the halotyrosine substrates, the series of zwitterion derivatives of tyrosine (X = H, Figure 2) were iodinated (X = I) prior to studies with a bacterial IYD (Scheme S1).

Figure 2.

Structures of tyrosine derivatives used to probe flavin activation in IYD. X = I, F, NO₂. X-HPPA, 3-(3-X-4-hydroxyphenyl)-propanoic acid; X-HPLA, (S) 3-(3-X-4-hydroxyphenyl)lactic acid.

Replacing the carboxylate of I-Tyr with a carboxamide group (I-tyrosinamide) removes the anionic charge but retains the potential for hydrogen bonding. This modification had little effect on affinity for IYD containing flavin in its oxidized and resting form (FMN_ox) (Table 1). Even the complete loss of the carboxylate group (I-tyramine) only diminished affinity by less than 18-fold. In contrast, loss of the α-ammonium group (3-(4-hydroxy-3-iodophenyl)propanoic acid, I-HPPA) decreased affinity by at least 700-fold and beyond a value that could be measured due to solubility limits of I-HPPA (Table 1). Substituting the α-ammonium group with an α-hydroxy group ((S)-3-(4-hydroxy-3-iodophenyl)lactic acid, I-HPLA) maintains hydrogen bonding, but the lack of charge was not sufficient to restore binding to IYD. Thus, coordination of the cationic ammonium group to the flavin O⁴ is critical for substrate coordination. This ammonium group also coordinates to a highly conserved carboxylate of an active site Glu, and conversion of this Glu to Gln in Drosophila IYD decreased the binding affinity of I-Tyr by 180-fold.²⁵ Surprisingly, substitution of a conserved Lys that coordinates to the substrate carboxylate with Gln had a greater impact on I-Tyr binding, as evident from a 370-fold loss of affinity.²⁵ This is contrary to the impact of losing the carboxylate from the substrate, and thus its zwitterion differentially interacts with the flavin and proximal amino acid side chains. Substrate recognition by wild-type IYD appears to demand the presence of an α-ammonium group on its substrates and is agnostic to the adjacent carboxylate, although both seem to interact with the flavin and the protein.

Table 1.

Binding and Deiodination of I-Tyr Zwitterionic Derivatives with IYD^a

compound	Kd (μM)b	kcat (s−1) × 10−2c	Km (μM)c	kcat/Km (M−1 s−1)
I-Tyrd	5.7 ± 0.4	60 ± 0.5	49 ± 10	(1.2 ± 0.2) × 104
I-tyrosinamide	7.0 ± 0.2	1.7 ± 0.2	2410 ± 410	5.7 ± 0.7
I-tyramine	99 ± 10	1.8 ± 0.3	1470 ± 450	13 ± 4.5
I-HPPA	>4000	<9.3 × 10−3e
I-HPLA	>3000	1.3 ± 0.1	270 ± 30	47 ± 5
D-I-Tyr	270 ± 15	2.7 ± 1.0	7900 ± 4000	3.4 ± 1

^aData represent the average of two independent measurements and the error represents their range.

^bSee Figure S1.

^cSee Figure S2.

^dFrom Kozyryev et al.²⁶

^eEstimated from the detection limit of HPPA by A₂₈₀ during HPLC analysis.

Despite the relatively mild loss of affinity for IYD caused by neutralization or deletion of the carboxylate group of I-Tyr, catalytic turnover was significantly impaired. For both I-tyrosinamide and I-tyramine, k_cat values for dehalogenation decreased by more than 30-fold and their overall efficiency of turnover (k_cat/K_m) decreased by more than 900-fold (Table 1). Thus, the substrate carboxylate group that coordinates to the flavin N3 helps to stabilize intermediates formed during catalysis but offers little stabilization to the resting state. This is consistent with earlier investigations that demonstrated a lack of correlation between the binding of halophenol derivatives and their efficiency of dehalogenation by IYD.¹⁹ The α-ammonium group may also be necessary for stabilizing reaction intermediates, but the lack of binding by I-HPPA limited our ability to measure dehalogenation. The k_cat value for catalysis can be no greater than 0.015% of that for I-Tyr, as set by our detection limit of the deiodinated product HPPA. While substitution of the α-ammonium group with an α-hydroxy group did not enhance binding to the resting enzyme, detectable catalytic activity was restored. I-HPLA was dehalogenated with a k_cat that was similar to the carboxylate derivatives, and its k_cat/K_m surpassed those of the carboxylate derivatives by more than threefold. Coordination of the zwitterion to both the flavin O⁴ and N3 are consequently essential for efficient reductive dehalogenation. Note that the loss of catalysis experienced by the substrates lacking the full zwitterionic structure is significantly greater than the 15-fold loss in k_cat/K_m caused by either (i) a loss of hydrogen bonding to the flavin N5 resulting from a Thr to Ala substitution or (ii) a loss of coordination to the flavin O² by a Lys to Gln substitution.^18,24 The catalytic efficiency for substrates with perturbations to their zwitterions is also quite similar to that for 2-iodophenol (k_cat/K_m = 0.7 M⁻¹ s⁻¹),¹⁹ which is devoid of these groups entirely. Together, these comparisons suggest that the α-ammonium and carboxylate groups of the substrate are most critical for activating the flavin for sequential single-electron transfer and reductive dehalogenation. Thus, IYD appears distinct from the vast majority of flavoproteins that rely on protein, rather than substrate, interactions to modulate the catalytic activity of their bound flavin cofactor.

Binding and turnover of the unnatural enantiomer d-3-iodotyrosine (d-I-Tyr) was also examined to test the spatial hierarchy of interactions established by the substrate zwitterion. In this example, electrostatic repulsion could be expected for interactions between d-I-Tyr and the flavin O⁴ and N3 groups as well as the Glu and Lys side chains described above. Instead, binding of d-I-Tyr to native IYD was moderately weakened but not nearly to the extent of that measured for I-HPPA or I-HPLA (Table 1). Perhaps the active site has a flexibility to accommodate the reorientation of charged groups in the substrate, but only a low basal level of catalysis was observed similar to that of the natural enantiomer of I-tyrosinamide. The values measured for d-I-Tyr are expected to reflect properties of this enantiomer rather than those arising from a trace impurity of the natural l-enantiomer. d-I-Tyr was treated with l-amino acid oxidase to remove any l-enantiomer that originated in the commercial starting material d-tyrosine (see the Supporting Information). Complementary analysis with l- and d-amino acid oxidases also indicated that d-I-Tyr had an enantiomeric purity of 99 ± 1% (Figures S3).

To probe the plasticity of the IYD active site further, the Glu and Lys that provide complementary interactions to the α-ammonium and carboxylate groups of (l-)I-Tyr were switched by site-directed mutagenesis to their E91K and K116E counterparts. This variant containing the double substitution expressed a diminished ability to discriminate between the substrate enantiomers based on K_d values for (l-) and d-I-Tyr of 85 ± 14 and 200 ± 30 μM, respectively (Figure S4). No catalytic dehalogenation could be detected for either enantiomer with this IYD variant, demonstrating once again that binding affinity is not predictive of catalytic activity. The placement and nature of the polar substrate and protein side chains are hence immutable for effective turnover.

Closure of the Active Site Lid Induced by Ligand Binding.

A correlation between the catalytic efficiency of IYD and closure of its active site lid was previously suggested by the contrasting catalytic and crystallographic information on I-Tyr and 2-iodophenol.¹⁹ More recent data based on engineering of the active site lid sequence and measurement of the dynamics of lid closure by NMR began to question the significance of this correlation.^22,23 The zwitterion derivatives described here offer an opportunity to clarify the role of lid closure for catalysis. Limited proteolysis has been a convenient tool for characterizing solvent accessibility of the lid.^32,42,43 In the absence of ligand, an Arg and a Lys within the lid are most facile targets of trypsin.²² These same residues gain protection from trypsin as the lid condenses around its substrates and ligands. IYD in the absence of ligand was consumed by trypsin (1:50, w:w) with a half-life of 10 min (Table 2 and Figures S5 and S6). Resistance to proteolysis increased over twofold in the presence of a saturating concentration of I-Tyr that is consistent with lid closure.²² This effect was not due to general inhibition of trypsin, as determined by its turnover of N-benzoyl-dl-arginine- 4-nitroanilide hydrochloride (dl-BAPNA, Figure S7). A slightly greater level of protection was gained from the presence of either I-tyrosinamide or I-tyramine (Table 2). These observations may not be surprising based on their reasonable affinity for IYD. Thus, active site binding was sufficient to secure the lid but not guarantee dehalogenation activity.

Table 2.

Half-Lives of IYD Proteolysis by Trypsin in the Presence and Absence of Iodinated Ligands^a

ligand	t1/2 (min)b	fold change in t1/2
none	10 ± 1
I-Tyr	22 ± 2	2.2
I-tyrosinamide	27 ± 2	2.7
I-tyramine	27 ± 8	2.7
d-I-Tyr	6.2 ± 1.2	0.62

^aValues and error were obtained from fitting two independent measurements to a first-order decay in Origin 6.0.

^bSee Figures S5 and S6.

d-I-Tyr is the only ligand to date that appears to promote trypsin-dependent hydrolysis of IYD relative to that in its absence. A similar activation by d-I-Tyr is not observed when dl-BAPNA is used as a substrate for trypsin and more likely d-I-Tyr is a mild inhibitor of trypsin (Table 2 and Figures S5–S7). Perhaps the aromatic ring of d-I-Tyr stacks over the flavin in analogy to (l-)I-Tyr and 2-iodophenol and orients the zwitterion in a manner that prevents lid closure and enhances its exposure relative to the dynamic range of conformations available in its absence.¹⁹ The E91K/K116E variant of IYD is somewhat more susceptible than the wild-type IYD to trypsin in the absence of ligands but gains over a twofold protection in the presence of either (l-)I-Tyr or d-I-Tyr (Table S1 and Figures S8 and S9). The protection gained in the presence of native (l-)I-Tyr suggests that active site binding and lid closure are maintained despite the lack of subsequent dehalogenation. The results with d-I-Tyr are more ambiguous since this enantiomer inhibits trypsin and may protect IYD through a nonspecific mechanism (Figure S7).

Substrate-Induced Stabilization of the FMN_sq.

In-efficient dehalogenation of substrates lacking the ability to coordinate with the flavin O⁴ and N3 is best explained by their inability to stabilize the intermediate FMN_sq that is critical for catalysis. While direct observation of FMN_sq during turnover can be difficult, prior investigations have utilized the binding of F-Tyr to IYD. This is an inert substrate analogue known to bind to the active site in analogy to I-Tyr and stabilize FMN_sq during reductive titration.^17,20 Accordingly, L-tyrosinamide and L-tyrosamine were fluorinated to prepare F-Tyr derivatives containing perturbations to the carboxylate group (see the Supporting Information). Equivalent fluorination of HPPA and HPLA was not pursued since neither of their iodinated derivatives were observed to bind IYD (Table 1). As expected from prior investigations,¹⁷ the affinity of F-Tyr for IYD (K_d = 97 ± 3 μM) is more than 10-fold weaker than that of I-Tyr (Table 1 and Figure S10). However, the trend in affinity after modification of its carboxylate group is similar to that of I-Tyr. F-tyrosinamide (X = F, Figure 2) exhibited a similar affinity (K_d = 77 ± 3 μM), as F-Tyr and F-tyramine affinity decreased by approximately 1 order of magnitude (K_d = 730 ± 140 μM) (Figure S10). Both protected IYD against trypsin as well, although the concentration necessary for saturating the active site with F-tyramine (7 mM) also inhibited trypsin (Figures S11–S13).

Standard reductive titration of IYD with xanthine and xanthine oxidase under anaerobic conditions generated the fully reduced flavin (FMN_hq) without a detectable FMN_sq intermediate (Figure 3). In contrast, an equivalent titration in the presence of saturating F-Tyr generated a new chromophore before full reduction with a broad absorbance around 600 nm that is consistent with the neutral form of FMN_sq and equivalent to that detected with IYD from mouse and human.^17,44 This species increases during the initial phase of titration and then decreases to yield the spectrum of FMN_hq (Figure 3). An equivalent formation of FMN_sq is not detected during titrations in the presence of saturating concentrations of either F-tyrosinamide or F-tyramine despite their ability to bind to the IYD active site and likely induce closure of the active site lid. While the carboxylate charge was not critical for binding of I- or F-tyrosinamide, this charge still controls the accumulation of FMN_sq. A similar lack of FMN_sq accumulation was observed when a hydrogen bond to the flavin N5 was disrupted in human IYD,^17,20 but surprisingly, this had much less impact on catalytic efficiency than the loss of the carboxylate group that interacts with the flavin N3. Coordination of F-tyrosinamide and F-tyramine also moderately destabilize the FMN_hq form of IYD, as evident by their respective two-electron potentials of −236 and −225 mV that are lower than the −200 mV measured in the absence of ligands (Table S2 and Figure S14). This effect is surprising since a net positive electrostatic potential would usually be expected to relieve destabilization of the reduced FMN_hq.⁴⁵ In comparison to perturbations at the flavin N3, disruption of the hydrogen bond to flavin N5 in human IYD destabilized FMN_hq to a lesser extent. In this case, the redox potential only decreased to −215 mV from the wild-type control of −200 mV.²⁰ Thus, the substrate zwitterion again appears to dominate the properties of the flavin in IYD.

Figure 3.

Redox titration of IYD in the presence of F-Tyr and its zwitterionic derivatives. IYD containing FMN_ox (20 μM) in 110 mM potassium phosphate pH 7.4 at 25 °C was reduced in the alternative presence of (A) no substrate, (B) F-Tyr (1 mM), (C) F-tyrosinamide (1 mM), and (D) F-tyramine (7 mM) using the xanthine/xanthine oxidase system.^20,28 Loss of absorbance at 450 nm corresponds to the reduction of FMN_ox, and an increase in absorbance at 600 nm corresponds to formation of FMN_sq.

Hydrogen bonding to the flavin N3 is common in flavoproteins and their model systems, but rarely is this interaction considered the key to stabilizing FMN_sq.^4,10 Flavodoxin from Clostridium beijerinckii represented a possible exception by orienting a protein carboxylate side chain of Glu next to the N3 of its bound flavin in analogy to the substrate carboxylate for IYD. Initially, the Glu was considered critical for FMN_sq formation based on its response to a Glu to Gln substitution. Further study reassigned the significance of the Glu from directly influencing N3 to indirectly positioning the FMN for maximizing the hydrogen bonding to its N5 position.^9,46 As typical, interactions to N5 once again remained dominant. The carboxylate provided by substrates of IYD are not likely to function merely as an anchor for orienting substrate since the affinity of I-tyrosinamide is not significantly perturbed relative to that of I-Tyr (Table 1). Hydrogen bonding to the O⁴ of FMN offers an additional site for controlling its redox properties.^13,47 Typically, interactions at this site are also established by the surrounding protein but for IYD these interactions derive from the α-ammonium group of substrate. Unfortunately, the significance of this group on the redox properties of FMN could not be characterized in IYD by the methods above since the ligands lacking the α-ammonium group had insufficient affinity for the active site. Hydrogen bonding to the O² group of flavin represents yet another site for redox control.¹² In IYD, an Arg coordinates to this region of FMN and remains constant in the presence and absence of ligand.^17,19 This interaction represents the only polar interactions provided by the protein. All others are supplied by the substrate, and thus stabilization of the intermediate FMN_sq and activation of catalysis depends on substrate structure.

Substrate-Induced Nitroreductase Activity of IYD.

If the substrate zwitterion is truly controlling the FMN chemistry to promote dehalogenation via a FMN_sq intermediate, then derivatives of the zwitterion have the potential to change the chemoselectivity of catalysis. Previously, destabilization of FMN_sq by loss of the hydrogen bond donor to the flavin N5 group suppressed dehalogenation and activated nitroreduction by IYD.¹⁸ Likewise, this same switch in catalytic specificity was observed in the absence of zwitterion interactions when studying turnover of 2-nitrophenol and 2-iodophenol.²⁰ At the time, the assumed lack of active site lid closure was used to rationalize the complementary loss and gain in catalytic function. Perturbations of the zwitterion of O₂N-Tyr offer an additional perspective for which binding and lid closure can be anticipated and nitroreductase activity may be inferred. To test these predictions, binding of the appropriate nitro compounds was characterized. IYD maintained a high affinity for both O₂N-Tyr and O₂N-tyrosinamide in analogy to I-Tyr and I-tyrosinamide (Tables 1 and 3 and Figure S15). In both examples, neutralization of the ligand carboxylate apparently had minimal impact on coordination to the N3 position of the resting-state FMN_ox. Coordination was more substantially perturbed by complete loss of the carboxylate group in O₂N-tyramine. In this case, binding weakened by 70-fold relative to that of O₂N-Tyr, well beyond the 17-fold loss for I-tyramine relative to that for I-Tyr. Unlike the iodinated derivatives, measurable binding was not lost upon substitution of the α-ammonium group with a hydroxy group. O₂N-HPLA bound to IYD with even greater affinity than O₂N-tyramine (Table 3 and Figure 2).

Table 3.

Binding of O₂N-Tyr Zwitterionic Derivatives to IYD^a

	A, B	Compound	kd(μM)b
	COO−, NH3+	O2N-Tyr	4.0 ± 0.9
	CONH2, NH3+	O2N-tyrosinamide	2.4 ± 0.1
	H, NH3+	O2N-tyraminc	280 ± 10
	COO−, OH	O2N-HPLA	180 ± 4

^aData represent the average of two independent measurements, and the error represents their range.

^bSee Figure S15.

Saturating concentrations of O₂N-tyrosinamide, O₂N-tyramine, and O₂N-HPLA all protected IYD from proteolysis by trypsin (Table 4 and Figures S16 and S17). For each, the half-life of IYD was extended by at least twofold over that in the absence of an active site ligand (Table 2). Thus, these ligands appear to secure the lid sequence surrounding the active site similarly to the analogous fluoro and iodo derivatives described above. At least for O₂N-tyrosinamide, protection of IYD was not the consequence of a general inhibition of trypsin as suggested by activity assays with DL-BAPNA as a substrate (Figure S18). Unfortunately, high concentrations required for saturation of the other ligands generated a very high background at A₄₀₅ that is used to detect the product of dl-BAPNA hydrolysis. For O₂N-HPLA, a concentration of only twice its K_d value could be examined with trypsin and these conditions resulted in no more than 25% inhibition of catalysis (Figure S18). With a confirmation that the nitro derivatives bound to IYD and secured closure of the lid, the nature of turnover could be examined.

Table 4.

Half-Lives of IYD Proteolysis by Trypsin Incubated with O₂N Ligands^a

ligand	t1/2 (min)b	fold increase in t1/2
O2N-tyrosinamide	30 ± 2	3.0
O2N-tyramine	27 ± 9	2.7
O2N-HPLA	21 ± 4	2.1

^aValues and error were obtained from fitting two independent measurements at each time point to a first-order decay in Origin 6.0.

^bSee Figures S16 and S17.

Nitroreductase activity was detected by the ability of substrates to oxidize the reduced form of IYD, as previously described.^18,21,48 IYD was first reduced by minimal dithionite, and then spectral changes were monitored after sequential addition of O₂N-Tyr or its derivatives. Equimolar concentrations of O₂N-Tyr were not capable of oxidizing IYD despite the potential for O₂N-Tyr to consume two hydride equivalents while generating its hydroxylamine product (Figure 4). These conditions oxidized FMN_hq by no more than approximately 30%, as evident from the gain in A₄₅₀, and simultaneously generated only an intermediate level of FMN_sq, as evident from the gain in A₆₀₀ that was also observed in earlier studies.¹⁸ In contrast, equimolar concentrations of O₂N-tyrosinamide, O₂N-tyramine, and O₂N-HPLA were each capable of oxidizing FMN_hq to regenerate the initial FMN_ox spectrum of IYD with a λ_max of ~450 nm, and this was accomplished with the latter two substrates at concentrations well below their K_d values (Table 3). The reciprocity between dehalogenation and nitroreductase can now be extended to substrates derivatized at their zwitterion. Loss of the carboxylate group by its conversion to a carboxamide analogue or its complete deletion destabilizes single-electron processes while concurrently stimulating a hydride transfer required for nitroreduction. Similarly, replacement of the α-ammonium group with a hydroxyl group activates hydride transfer and suppresses single-electron transfer (Figure 4 and Table 1). The polar interactions between the zwitterion and the N3 and O⁴ groups of flavin are thus key for supporting efficient dehalogenation by IYD. The absence of either the substrate α-ammonium or carboxylate group impacts catalysis to a far greater extent than loss of a hydrogen bond to its N5 position.

Figure 4.

Oxidation of FMN_hq with O₂N-Tyr and its zwitterionic derivatives. (A) IYD (oxidized, 18 μM, black trace) in 110 mM potassium phosphate pH 7.4 was reduced with a minimal amount of dithionite (red trace) under anaerobic conditions. This was confirmed to lack excess dithionite by adding 20 μL of air-saturated buffer and observing partial oxidation of IYD. Aliquots of 6 μM O₂N-Tyr (green, blue, and cyan traces) were added sequentially at 30 min intervals for a final concentration of 18 μM. Independent samples of IYD (black traces, 15 μM) were reduced under equivalent conditions, and two aliquots each of 5 μM of (B) O₂N-tyrosinamide, (C) O₂N-tyramine, and (D) O₂N-HPLA were alternatively added for final concentrations of 10 μM (green and blue traces, respectively).

The significance of polar interactions to the N3 and O⁴ groups of flavin serves IYD well. Catalysis is controlled by direct coordination between the substrate zwitterion and flavin to limit dehalogenation to halotyrosines at the amino acid level and avoid processing these residues embedded in peptides or proteins. At least for vertebrates, this mechanism of activation safeguards the iodotyrosyl residues that accumulate on the surface of thyroglobulin as intermediates during formation of thyroid hormone. The advantage for this same specificity in bacteria IYD as reported here will not be obvious until the physiological role of this enzyme can be identified. Whether a comparable substrate of bromotyrosine discovered in fruit flies is generated as an amino acid or derives from bromotyrosine containing proteins or peptides remains to be determined as well.⁴⁹

When surveying conserved residues in the nitroreductase superfamily, structure−function relationships are most clearly established by residues adjacent to the flavin N5 and much less at the O², N3, and O⁴.¹⁴ Consequently, the significance of the substrate zwitterion was unanticipated but now explains a variety of prior results including the relatively small loss in dehalogenase activity when hydrogen bonding to flavin N5 was removed.¹⁸ This discovery also explains why only a modest increase in dehalogenation of 2-iodophenol was achieved after redesigning the active site lid for stabilizing the association of this substrate with IYD.²³ Future efforts for such engineering will require the additional positioning of side chains to establish polar interactions with the N3 and O⁴ of flavin. The mere proximity created by stacking of a phenolic derivative over FMN_hq is not sufficient for single-electron transfer and dehalogenation. Stabilization of the FMN_sq intermediate is also necessary whether this is supported by substrate or protein. In contrast, stacking and orientation may be sufficient to promote hydride transfer for nitroreduction by IYD.

CONCLUSIONS

The broad range of reactions promoted by flavin cofactors provide biology with great opportunities but also significant risks. The chemistry of this cofactor requires exquisite control to ensure a high specificity for the desired catalysis and protection against undesirable and often detrimental processes. Typically, such control is the responsibility of the surrounding protein that variably employs sterics, solvation, electrostatics, hydrogen bonding, acid–base chemistry, and proximity to manipulate the many modes of flavin reactivity. IYD offers one of the few examples of a flavoprotein that establishes direct coordination between polar functional groups of the substrate and flavin. Initially, these features were thought to help template active site organization and lid closure to align a Thr side chain for hydrogen bonding to the all-important flavin N5 position.¹⁷ This was further supported by crystallographic and kinetic studies connecting lid closure with efficient dehalogenation and lid disorder with a significant decrease in dehalogenation and increase in nitroreduction.^19,20 However, solution studies by NMR revealed significant closure of the lid even with phenolic ligands and thus an alternative basis of FMN activation was necessary.²² The zwitterion ion derivatives of I-Tyr, F-Tyr, and O₂N-Tyr have all now generated data consistent with a control of catalytic specificity that is dominated by interactions between the substrate zwitterion and the N3 and O⁴ groups of flavin. Although active site binding tolerates modifications of the substrate’s carboxylate group, efficient dehalogenation requires the presence of both the α-ammoniun and carboxylate groups. Together, these stabilize the FMN_sq required for a mechanism of dehalogenation involving single-electron transfer. Such interactions are not required for a basal nitroreductase activity that is apparent only when formation of FMN_sq is suppressed. The nature of the substrate consequently dictates the chemistry of catalysis supported by IYD in a manner that is highly distinct from most other flavoproteins.