Strongly Coupled Redox-Linked Conformational Switching at the Active Site of the Non-Heme Iron-Dependent Dioxygenase, TauD

Abstract

2-Oxoglutarate (2OG)-dependent dioxygenases catalyze C─H activation while performing a wide range of chemical transformations. In contrast to their heme analogues, non-heme iron centers afford greater structural flexibility with important implications for their diverse catalytic mechanisms. We characterize an in situ structural model of the putative transient ferric intermediate of 2OG:taurine dioxygenase (TauD) by using a combination of spectroelectrochemical and semiempirical computational methods, demonstrating that the Fe(III/II) transition involves a substantial, fully reversible, redox-linked conformational change at the active site. This rearrangement alters the apparent redox potential of the active site between −127 mV for reduction of the ferric state and +171 mV for oxidation of the ferrous state of the 2OG-Fe-TauD complex. Structural perturbations exhibit limited sensitivity to mediator concentrations and potential pulse duration. Similar changes were observed in the Fe-TauD and taurine-2OG-Fe-TauD complexes, thus attributing the reorganization to the protein moiety rather than the cosubstrates. Redox-difference infrared spectra indicate a reorganization of the protein backbone in addition to the involvement of carboxylate and histidine ligands. Quantitative modeling of the transient redox response using two alternative reaction schemes across a variety of experimental conditions strongly supports the proposal for intrinsic protein reorganization as the origin of the experimental observations.

Graphical Abstract

INTRODUCTION

The 2-oxoglutarate (2OG)-dependent dioxygenases activate C─H bonds while catalyzing a variety of biologically relevant reactions,¹ including synthesis of a wide range of commercial products.² 2OG:taurine dioxygenase (TauD), the archetypical member of this enzyme family^3,4 is found in Escherichia coli where it metabolizes taurine (2-aminoethanesulfonate) as a sulfur source for sulfur-starved cells.⁵ TauD activates taurine via hydrogen atom transfer (HAT) to an Fe(IV)═O intermediate J (shown as F4 in Figure 1),⁶ followed by substrate oxygenation and product degradation.⁷

Figure 1.

Redox states of catalytic intermediates and artificial complexes of TauD. Vertical transitions indicate a change in the oxidation state of the Fe center. Diagonal transitions indicate a change in a protonation state. Gray structures show the classical hydroxyl radical rebound pathway. Blue structures illustrate electrochemical manipulation of TauD to model the F3 state.

A time-resolved resonance Raman study of TauD revealed the existence of two transient intermediates that trailed F4 in time and assigned them to the ν_Fe–O and ν_as modes of the ferric (hydr)oxo (F3) and alkoxo (FX) intermediates, respectively.⁸ HAT by F4 in TauD is analogous to HAT by compound I in cytochromes P450 (CYP450)⁹ except that TauD is expected to yield a transient non-heme Fe(III)─OH complex (Figure 1) instead of a heme Fe(IV)─OH in compound II of CYP450. However, the Raman study found no vibrational evidence for protonation of the oxo group in F3, attributing its absence to a proton transfer to a nearby base that allows for the formation of an ensuing FX species. While the alkoxo structure of FX received additional support in a recent crystallographic study of another 2OG-dependent hydroxylase, VioC,¹⁰ the lack of proton sensitivity of F3 is unexpected for two reasons. First, the Raman study⁸ was conducted at pH 8 and well below the pK_a ≈ 25 (in DMSO) reported for Fe(III)─OH model complexes.¹¹ The low dielectric environment of the active site¹² is also expected to increase the pK_a over that in an aqueous medium. Second, the pK_a of F3 contributes to the HAT capacity of F4 per Bordwell’s thermodynamic cycle (Figure 1),^13,14 as has been demonstrated for the analogous pK_a of compound II in CYP450 (pK_a = 12)¹⁵ in contrast to peroxidases (pK_a < 4).^16,17 Since the geometry of the Fe(III)─OH ligand may reduce the sensitivity of the ν_Fe─OH Raman mode to ¹H/²H substitution, it is important to examine the protonation state of F3 using other techniques.

Spectroelectrochemistry can reveal protonation events either directly from spectroscopic changes or indirectly from the pH sensitivity of the Fe(II)/Fe(III) redox potential.^3,18,19 This information can then be used to reconstruct the protonation states of Fe(III)-TauD as an in situ model of the transient F3 species from the known crystallographic structures of Fe(II)-TauD. The weak UV–vis absorption of the non-heme iron center in TauD²⁰ makes direct detection of the Fe(II)/Fe(III) redox transition by optical spectroscopy impractical. Much of the pioneering work by Mäantele et al.^21-23 focused on proteins with porphyrin cofactors, with their strong, oxidation state-sensitive optical absorption spectra. Similar methods were applied to the relatively intense chromophores in copper enzymes^18,24 and Ni–Fe hydrogenases.^25,26 As an alternative to optical spectroscopy, one can exploit reaction-induced infrared (IR)-difference spectrocopy^27,28 for the detection of reversible vibrational changes of ligands in response to the redox state changes of the metal.^29,30 Here, we investigated several complexes of TauD (Figure 1) using normal pulse spectrovoltametry (NPSV) with IR detection to identify structural and redox properties simultaneously.³¹ This study revealed a redox-linked conformational change at the active site of TauD that modulates its redox potential at an unprecedented magnitude.

METHODS

Sample Preparation.

TauD apoprotein was purified as previously described³² with the following modifications: cell cultures were grown in six 2 L flasks, each containing 1 L of Terrific Broth medium, at 37 °C and with shaking at 200 rpm. A final concentration of 1 mM isopropyl β-d-1-thiogalactopyr-anoside was added to each flask after reaching an optical density of 0.6 at 600 nm. The temperature was reduced to 30 °C, and the cultures were grown overnight. Pelleted cells were stored at −80 °C until needed. The cells were thawed and resuspended in a lysis buffer containing 20 mM Tris, pH 8, 1 mM ethylenediaminetetraacetic acid and 1 mM phenylmethanesulfonyl fluoride prior to being lysed by sonication.

Fe(III)-TauD was prepared by adding a 1:1 molar ratio of ferrous ammonium sulfate to TauD apoprotein under anaerobic conditions, followed by oxidation using a 5-fold excess of ferricyanide (FCN) for 1 h and the removal of ferri/ ferrocyanide using a GE Healthcare PD-10 desalting column. Fe(III)-TauD samples were stored on ice in 25 mM Tris buffer, pH 8.

The final sample buffer was exchanged using a 10 kDa centrifugation membrane unit (Amicon). 2OG-Fe(III)-TauD and taurine-2OG-Fe(III)-TauD were prepared by adding 2-fold excess of 2OG and/or 3-fold excess of taurine to approximately 1 mM Fe(III)-TauD. FCN, methylene green (MG), and thionine acetate (TA) mediators were added to achieve 100 μM of each in the final solution, unless noted otherwise. All samples were prepared in 25 mM Tris buffer at pD 8.5 in D₂O with 0.5 M KCl.

Spectroscopic Measurements.

NPSV measurements were performed using a 12.5 μm path length optically transparent thin layer electrochemical (OTTLE) cell³¹ over a boron-doped diamond working electrode and a Ag/AgCl (0.5 M KCl) reference electrode at 10 °C.^33,34 Electrochemical measurements were performed using a computer-controlled potentiostat (Model CHI1202b, CH Instruments). Reference potentials (E_r) in reduction and oxidation modes were +0.5 and −0.6 V, respectively, with a potential increment of 0.04 V. The potential pulse duration was 300 s, unless otherwise noted, with FTIR spectral acquisition (Equinox 55/S, Bruker) during the final 120 s (Figure S1). NPSV data were analyzed by a nonlinear global spectral regression (GSR)³⁵ and kinetic simulations were performed using KinESim³⁶ packages for Igor Pro.³¹

The solution rate constants k_sol for the reaction of MG and TA with Fe(III)-TauD were determined as previously described for myoglobin (Mb).³¹ The k_sol for the oxidation of Fe(II)-TauD by FCN was determined anaerobically while stirring against 100 μM FCN in 25 mM Tris buffer, pH 8. The reaction was initiated by the injection of Fe(II)-TauD to achieve an equimolar ratio with the mediator. UV–vis spectra (Hewlett-Packard 8453) were collected every 2 s for 20 s before and 10 min after the start of the reaction. The bimolecular rate constant was obtained by fitting the temporal changes at 420 nm

RESULTS

NPSV (Figure S1) utilizes a series of applied potentials, E_a,i, each incremented by a defined potential step and preceded by a reference potential, E_r,i. The absolute Fourier transform infrared (FTIR) spectra were acquired at the end of each potential step over a spectral integration period when the reaction had reached equilibrium. Spectra acquired sequentially under alternating potentials E_a,i,(S_a,i) and E_r,i(S_r,i) formed a raw absolute data matrix (Figure S2, top). This matrix was reduced into a redox-difference matrix of spectra (ΔS_i, Figure S2, bottom):

$$\Delta {\mathbf{S}}_i={\mathbf{S}}_{\mathrm{a},i}-{}^1∕_2({\mathbf{S}}_{\mathrm{r},i}+{\mathbf{S}}_{\mathrm{r},(i+1)})$$ (1)

Experimental values of ΔS_i at 1681 cm⁻¹ are shown by markers in Figure 2, right. For the oxidation NPSV segment (blue) ΔA₁₆₈₁ was measured vs E_r,i = −0.6 V as E_a,i was increased from −0.14 to +0.34 V. For the reduction sweep (red), E_a,i was decreased from +0.14 to −0.38 V and ΔA₁₆₈₁ was measured vs E_r,i = +0.5 V.

Figure 2.

NPSV transitions in 2OG-Fe-TauD. Redox mediators: 100 μM MG, 100 μM TA, and 100 μM FCN. Left: ΔS_tot of reduction (red, vs E_r = 0.5 V) and oxidation (blue, vs E_r = −0.6 V) steps. Right: Experimental profiles of reduction steps (red circles) and oxidation steps (blue circles) normalized using ΔA₁₆₈₁ with ϕ_Rd (red) and ϕ_Ox (blue) profiles. For ease of comparison, both ΔS_tot and ϕ are shown as the Fe(III)/Fe(II) redox difference.

The redox-difference data set (ΔS) was deconvoluted into a full occupancy spectra, ΔS_tot, and Nernstian population profiles, ϕ, so that each redox-difference spectrum ΔS_i at the ith NPSV step is described as

$$\Delta {\mathbf{S}}_i=\Delta {\mathbf{S}}_{\text{tot}}\times {\varphi }_i$$ (2)

where ϕ_i = f(E_a,i,E_1/2,n) is a property of the analyte and the applied potential (E_a).³¹ The resulting profiles ϕ (lines in Figure 2, right) represent E_a-dependent intensities of the entire spectra ΔS_tot (Figure 2, left) as opposed to experimental ΔA at a selected frequency (markers).

Fe-TauD is redox inactive on the electrode without mediators (MG, TA, and FCN), each of which is also described by its own ΔS_tot, E_1/2, and n.³¹ Figure 2 shows ΔS_tot (left) and ϕ (right) of the Fe(II)/Fe(III) transition in 2OG-Fe-TauD, while Figure S3 compares it with the same transition in Fe-TauD and taurine-2OG-Fe-TauD. Notably, mediators are described by their own matrices ΔS = ΔS_tot × ϕ and do not contribute to TauD spectra.

The close similarities between the reduction and oxidation FTIR spectra in all three complexes show that the observed redox processes are fully reversible. The spectra were dominated by vibrational changes in the 1550–1700 cm⁻¹ region, outside of which only a vibration at ~1400 cm⁻¹ was consistently observed in all forms. Redox changes in the 1650–1700 cm⁻¹ region were specific to the 2OG-containing complexes and were altered upon taurine binding (Figure S3). All complexes exhibited changes in the amide I stretching region at 1632/1624 cm⁻¹ (Fe-TauD) or 1638/1630 cm⁻¹ (2OG-Fe-TauD and taurine-2OG-Fe-TauD).³⁷ A prominent trough was observed at 1580 cm⁻¹ in Fe-TauD with a likely corresponding peak at 1614 cm⁻¹; these features showed a 20 cm⁻¹ downshift upon binding of 2OG (2OG-Fe-TauD and taurine-2OG-TauD).

In contrast to the redox spectra, the reduction and oxidation population profiles, ϕ_Rd and ϕ_Ox, were sharply different, exhibiting a large separation of apparent potentials (Figure 2, right) and the appearance of minor oxidation phases in Fe-TauD and taurine-2OG-Fe-TauD (Figure S3). Results of fitting ϕ to Nernstian profiles with n = 1 are shown in Table S1. The appearance of a large redox hysteresis in all three forms of TauD contrasts with the behavior of Mb using the same OTTLE cell, electrode, and mediators (except FCN),³¹ raising the possibility that the shift in potentials originates from the difference in the specific interactions of reduced and oxidized forms of the mediator with the protein moiety rather than $E_{1∕2}^{\mathrm{A}}$ of the metal. This possibility was excluded using Zn²⁺-substituted TauD,²⁰ which was found to be redox inactive (Figure S4), attributing the observed changes to the redox transitions of the iron center.

The second possibility is that the apparent shift in potentials arises from the kinetic or thermodynamic limitations of the TauD/mediator reactions, particularly due to the redox gap between the E_1/2 of FCN and the other two mediators (Table S2), which is comparable in magnitude to the hysteresis observed in TauD. Since this potential gap could not be bridged experimentally due to low solubility, instability, or irreversible interactions of other mediators with TauD (Table S3), the role of kinetic limitations of a discontinuous ladder of mediators was investigated using a semiempirical model of heterogeneous mediated electrochemistry (Figure S5).³¹ Since this model was developed using the same mediators and electrochemical system as reported here, it required only parametrization of FCN, including the mediator-specific heterogeneous potential-dependent (k*_el) and potential-independent (k_lim) effective rate-limiting constants on the electrode and the bimolecular rate constant (k_sol) for the homogeneous FCN/TauD reaction (Table S2). These kinetic parameters were used to simulate concentration profiles of TauD, which were integrated over specified time periods to obtain ϕ and compared with experimental profiles obtained under identical conditions.

Two alternative reaction schemes were investigated to assess the origin of the redox hysteresis in TauD (Figure 3). In the single state model [1], identical to what was used for Mb,³¹ TauD exists in a simple redox equilibrium with a single redox potential, $E_{1∕2}^{\mathrm{A}}$. The redox hysteresis using [1] could originate only from the inefficiency of the mediators. The redox-linked switching model [2] includes at least two distinct conformations of TauD, conformers A and A′, each with a distinct redox potential (Figure 3, right). Since the experimental NPSV data report the combined contribution of all isomers of TauD, simulated populations of conformers A and A′ were combined as well. The combined populations yielded only one or two apparent NPSV transitions even if the responses of individual conformers were more complex. Conformational changes of TauD using [2] are described by separate sets of rates and equilibria constants in the reduced and oxidized states. The values of k₁ and k₂ were estimated from the slow phases of oxidation by FCN and were found to be approximately 9 × 10⁻⁴ s⁻¹ (Figure S6). The extent of the oxidation of Fe(II)-TauD by FCN over the first 30 s allowed for the estimation of K₁ ≥ 10² and the reverse isomerization rate k₋₁ ≈ 10⁻⁵ s⁻¹. Since ΔG = 0 for a cyclic process, one can calculate K₂ = 9.6 × 10⁻⁶ from experimental estimates $E_{1∕2}^{\mathrm{A}}$, $E_{1∕2}^{{\mathrm{A}}^{\prime }}$, and K₁.

Figure 3.

Chemical models for the mediated electrochemistry of TauD. The single state model [1] describes a homogeneous population of A, with a single redox potential, $E_{1∕2}^{\mathrm{A}}$. The redox switching model [2] allows for redox-linked isomerization of the analyte between conformations A and A′ with distinct redox potentials, $E_{1∕2}^{\mathrm{A}}$ and $E_{1∕2}^{{\mathrm{A}}^{\prime }}$, respectively, and preferential stability in the oxidized and reduced forms. Direct electron transfer is allowed only for the mediator M. See text for the estimates and interpretation of forward and reverse isomerization rates.

The NPSV response was predicted for the two alternative models for 2OG-Fe-TauD while varying $E_{1∕2}^{\mathrm{A}}$ (for [1]) or $E_{1∕2}^{\mathrm{A}}$ and $E_{1∕2}^{{\mathrm{A}}^{\prime }}$ (for [2]). The simulated ${\varphi }_{\mathrm{Ox}}^{\mathrm{A}}$ and ${\varphi }_{\mathrm{Rd}}^{\mathrm{A}}$ profiles shown in Figure 4 and Figure S7 are directly comparable to those obtained from experimental NPSV data in Figure 2, right. Simulations using [1] with an $E_{1∕2}^{\mathrm{A}}$ within 50 mV of that of any mediator show nearly complete reversibility, as was observed for Mb.³¹ As $E_{1∕2}^{\mathrm{A}}$ approached the middle of the potential gap between MG and FCN, ϕ^A exhibited an increasing degree of distortion with the development of an apparent hysteresis between −60 and +140 mV in the reduction and oxidation modes, respectively, and the appearance of minor redox transitions (Figure 5 and Table S4). At $E_{1∕2}^{\mathrm{A}}=0$ mV the hysteresis reaches a maximum amplitude of 213 mV, which is much smaller than the 298 mV hysteresis observed in TauD. The largest hysteresis of 199 mV was observed using [1] with symmetrical ${\varphi }_{\mathrm{Ox}}^{\mathrm{A}}$ and ${\varphi }_{\mathrm{Rd}}^{\mathrm{A}}$ and $E_{1∕2}^{\mathrm{A}}=60\mathrm{mV}$ (Figure S7).

Figure 4.

Effect of thermodynamic properties of the analyte on the apparent NPSV redox hysteresis for models [1] and [2]. Simulated ϕ_Rd (red) and ϕ_Ox (blue) profiles of an analyte (A) that is completely dependent on mediators (Table S2). The intrinsic E_1/2 of A (left) or A, A′ (right) used in simulations are indicated to the right of each plot. The observed E_1/2 values are indicated next to each transition, including minor transitions, where present.

Figure 5.

Effect of the intrinsic E_1/2 on NPSV redox transitions in model [1]. Simulations and NPSV integrations were performed under realistic conditions. Marker sizes represent relative amplitudes of the major and minor fitted phases (E_Obs) of corresponding ϕ^A profiles in the reduction (red circle) and oxidation (blue circle) NPSV modes. The intrinsic $E_{1∕2}^{\mathrm{A}}$ illustrated in Figure 4, left, are indicated by vertical dashed lines and their intercepts with the plot represent the apparent transitions. The diagonal dashed line represents ideal Nernstian behavior. Arrows indicate the E_1/2 of individual mediators.

Analogous simulations using model [2], with the addition of $E_{1∕2}^{{\mathrm{A}}^{\prime }}$, confirmed that the profile of ϕ^TauD and the magnitude of the observed hysteresis depend primarily on the intrinsic potentials $E_{1∕2}^{\mathrm{A}}$ and $E_{1∕2}^{{\mathrm{A}}^{\prime }}$ (Figure 4) when their difference is large. When $E_{1∕2}^{\mathrm{A}}\approx E_{1∕2}^{{\mathrm{A}}^{\prime }}$, [2] reduces to [1] and the observed ϕ^TauD was determined by the gap in the mediator ladder. Therefore, there were multiple conditions where the shape of ϕ^TauD was not sufficient to unambiguously distinguish between [1] and [2] based on the hysteresis alone. However, the two models predict different sensitivities to mediator concentrations, particularly that of FCN due to the absence other mediators in its effective potential range.

The effects of FCN concentration on ϕ_Ox and ϕ_Rd using models [1] vs [2] are illustrated in Figure 6. For [1], we selected E_1/2 = 60 mV for which the hysteresis is the most sensitive to FCN concentration. Simulations using [2] were conducted with $E_{1∕2}^{\mathrm{A}}=-140\mathrm{mV}$ and $E_{1∕2}^{{\mathrm{A}}^{\prime }}=100\mathrm{mV}$, as these values yielded an apparent hysteresis comparable to that observed experimentally. An increase in the concentration of FCN from 0.1 mM to 2.5 mM decreased the oxidation E_Obs by 66 mV using [1] (131–65.2 mV) and 78 mV using [2] (158–80 mV) vs the experimentally observed decrease of 121 Mv (171–50 mV). Oxidation E_Obs using [2] decreased below $E_{1∕2}^{{\mathrm{A}}^{\prime }}$ in agreement with the continuous changes in experimental value (Figure 6, middle) and contrary to the exponential saturation predicted for [1]. However, the largest discrepancy was observed in the apparent reduction potential, which increased using [1] and remained essentially unchanged using [2]. As a result, [1] predicts the loss of the hysteresis at 2.5 mM FCN, in contrast to the hysteresis of 198 mV observed experimentally and 184 mV predicted using [2]. The contribution of the minor phase also diminishes at high FCN concentration.

Figure 6.

Effect of FCN concentration on the observed ϕ_Rd and ϕ_Ox profiles. Experimental ${\varphi }_{\mathrm{Rd}}^{\mathrm{TauD}}$ (red circle) and ${\varphi }_{\mathrm{Ox}}^{\mathrm{TauD}}$ (blue circle) for 1 mM TauD (center) are compared with corresponding profiles predicted by models [1] (left) and [2] (right) at the indicated concentrations of FCN. The $E_{1∕2}^{\mathrm{A}}$ using [1] was 60 mV. $E_{1∕2}^{\mathrm{A}}$ and $E_{1∕2}^{{\mathrm{A}}^{\prime }}$ using [2] were −130 and +100 mV, respectively. The E_Obs (mV) of major phases are shown.

The complete removal of FCN is expected to hinder the oxidation process and exacerbate the nonideal behavior of TauD differently for models [1] and [2], allowing for further discrimination between the models. As the applied potentials E_a,4, and E_a,5 approach E_1/2 (Figure 7, left; NPV cycles 4 and 5), facile reduction of the analyte leads to large changes in its population during the ${\overline{\varphi }}_{\mathrm{r},4}\to {\overline{\varphi }}_{\mathrm{a},4}$ and ${\overline{\varphi }}_{\mathrm{r},5}\to {\overline{\varphi }}_{\mathrm{a},5}$ steps. This effect increases the reversible NPSV response using [1] ($\Delta {\overline{\varphi }}_4$ and $\Delta {\overline{\varphi }}_5$), which is defined as $\Delta {\overline{\varphi }}_i={\overline{\varphi }}_{\mathrm{a},i}-0.5({\overline{\varphi }}_{\mathrm{r},i}+{\overline{\varphi }}_{\mathrm{a},i+1})$.³¹ However, this reduction is followed by only a minimal reoxidation during ${\overline{\varphi }}_{\mathrm{a},4}\to {\overline{\varphi }}_{\mathrm{r},5}$ and ${\overline{\varphi }}_{\mathrm{a},5}\to {\overline{\varphi }}_{\mathrm{r},6}$ steps due to the limited oxidizing capacity of low potential mediators. Subsequent steps show exhaustive reduction with smaller amplitudes (${\overline{\varphi }}_{\mathrm{a},6}$) and the NPSV response of the mostly reduced sample falls ($\Delta {\overline{\varphi }}_6$; Figure 7, right). This process results in a pseudoequilibrium region where changes in ${\overline{\varphi }}_i$ become potential-independent due to inefficient oxidation of TauD by MG and TA even at a high E_r (shaded area in Figure 7). The magnitude of $\Delta \overline{\varphi }$ in this region represents only the extent of the change in redox state ($\Delta {\overline{\varphi }}_7$ < 20%, right) and not the major redox state of TauD (>80% reduced ${\overline{\varphi }}_{\mathrm{a},7}$ and ${\overline{\varphi }}_{\mathrm{r},7}$, left).

Figure 7.

Kinetically limited NPSV reduction profile in the absence of FCN. Left: Concentration of oxidized analyte (solid line, E_1/2 = 80 mV) and oxidized MG (dashed line) predicted by [1] over successive NPSV cycles (dotted line). Simulation conditions for a realistic mediator cocktail are identical to the 1000 s pulse width data in Figure 8 (left). TA is present, but not shown for clarity. Integrated ${\overline{\varphi }}_i$ that determine $\Delta {\overline{\varphi }}_5$ and $\Delta {\overline{\varphi }}_7$ are shown in blue and red, respectively. Right: Integrated NPSV profile for Ea decreasing from +0.1 V to −0.2 V over eight steps against E_r = 0.1 V. The maximal amplitude is observed for $\Delta {\overline{\varphi }}_5$ due to facile reduction during the ${\overline{\varphi }}_{\mathrm{r},5}\to {\overline{\varphi }}_{\mathrm{a},5}$ step. The amplitudes of $\Delta {\overline{\varphi }}_{\ge 6}$ are decreased due to the slow reduction, i.e. ${\overline{\varphi }}_{\mathrm{a},7}\to {\overline{\varphi }}_{\mathrm{r},8}$ step. An expanded view of this profile is shown in Figure 8 (left top open markers) in comparison with shorter pulse duration and the oxidation mode profiles. The pseudoequilibrium region is highlighted in gray.

A different response can be expected using [2] in the absence of FCN, because protein isomerization provides an alternative pathway for oxidation of A′ via A. Furthermore, both models predict that nonequilibrium ϕ profiles would exhibit substantial dependence on pulse duration and analyte potentials. Predictions of the two models (Figure 8 left and right) differ in three characteristic parameters. (i) The onsets of the ϕ_Ox and ϕ_Rd transitions using [1] are narrow (up to $\Delta {\overline{\varphi }}_5$ in Figure 7) and follow the profiles of an n = 1 redox process. Using [1], this is followed by a sharp transition to the pseudoequilibrium phase ($\Delta {\overline{\varphi }}_6$ in Figure 7). In contrast, the onset is much broader using [2] (n = 0.2–0.5, for the best fit lines in Figure 8, right) and no distinct pseudoequilibrium region is observed. (ii) The alternative oxidation pathway using [2] results in a larger NPSV amplitude following the onset of the redox transition. A second NPSV peak in the reduction mode is predicted using [2] under some conditions (Figure S8), which reflects a direct contribution of the low potential redox transition. The pseudoequilibrium region using [1] is always featureless and its amplitude diminishes rapidly at high E_1/2. (iii) The apparent potentials of maximal ϕ_Ox and ϕ_Rd responses are close to the intrinsic $E_{1∕2}^{\mathrm{A}}$ using [1] (ΔE_Obs = 110 mV) in contrast to a much larger apparent difference for [2] (ΔE_Obs = 330 mV) (Figure 8, black lines).

Figure 8.

Comparison of experimental NPSV hysteresis with model-dependent predictions in the absence of FCN. Experimental ${\varphi }_{\mathrm{Rd}}^{\mathrm{TauD}}$ (red circles) and ${\varphi }_{\mathrm{Ox}}^{\mathrm{TauD}}$ (blue circles) of 1 mM TauD with 0.5 mM MG and saturated TA (<150 μM) but no FCN (center) are compared with corresponding profiles predicted by models [1] (left) and [2] (right). The origin of the characteristic reduction response using [1] (top left) is illustrated in Figure 7. Experimental and simulated profiles were acquired for 300 s (solid circles) and 1000 s (open circle) NPSV pulse widths. Best fit Nernstian profiles are shown for reference (—; [1], n = 1, ΔE_obs = 0.11 V; [2], n_ox = 0.53, n_rd = 0.23, ΔE_obs = 0.33 V).

The experimental ϕ^TauD, obtained at 300 and 1000 s NPSV pulse widths, are compared to the matching simulations using models [1] and [2] in Figure 8 and S7. As argued above for Figure 4, experimental observations for the full mediator cocktail show that $E_{1∕2}^{\mathrm{A}}$ using [1] is limited to the range of 40–80 mV, but these $E_{1∕2}^{\mathrm{A}}$ values result in a substantially larger ϕ amplitude in the absence of FCN than observed experimentally (Figure S8). The maximum experimental NPSV amplitude of 2OG-Fe-TauD with 300 and 1000 s pulses in the absence of FCN (Figure 8) was >0.7 and <0.18 mM for the reduction and oxidation modes, respectively. Such an amplitude was observed using [1] only for $E_{1∕2}^{\mathrm{A}}$ of +100 to +130 mV, where the hysteresis in the full mediator cocktail is already smaller than observed for TauD. The experimental NPSV onset was much broader than predicted for [1] with E_1/2 < +130 mV and showed a better correlation using simulation [2] with $E_{1∕2}^{{\mathrm{A}}^{\prime }}>+140\mathrm{mV}$. The relative amplitudes of ${\varphi }_{\mathrm{Ox}}^{\mathrm{TauD}}$ and ${\varphi }_{\mathrm{Rd}}^{\mathrm{TauD}}$ with 300 or 1000 s NPV pulse widths further support model [2] over [1].

DISCUSSION

The redox cycle of anaerobic Fe-TauD is vibrationally fully reversible (Figure 2). The observation of an electrochemical hysteresis in all forms of TauD (Figure S3) was surprising and sharply contrasts with the simple redox transition of myoglobin.³¹ The lack of suitable mediators with E_1/2 of 0–100 mV required us to conduct an investigation into the role of kinetic and thermodynamic limitations of mediated electrochemistry using two alternative models (Figure 3) before being able to state unequivocally that the hysteresis is attributed to intrinsic properties of Fe-TauD. The application of a semiempirical model of heterogeneous thin layer electrochemistry showed that both models can yield comparable NPSV results using specific Fe-TauD potentials at the given mediator concentrations. However, no single $E_{1∕2}^{\mathrm{TauD}}$ value in model [1] could yield ϕ_Ox and ϕ_Rd consistent with experimental observations across all the conditions tested here, giving strong support for intrinsic redox-linked reorganization in Fe-TauD, as follows.

The largest hysteresis predicted for model [1] was 213 mV ($E_{1∕2}^{\mathrm{TauD}}=0.0\mathrm{mV}$), which is 85 mV smaller than that experimentally observed in 2OG-Fe-TauD and taurine-2OG-Fe-TauD and 128 mV smaller than that observed in Fe-TauD (Table S1). The magnitude of the hysteresis and apparent $E_{1∕2}^{\mathrm{TauD}}$ values in 2OG-Fe-TauD and taurine-2OG-Fe-TauD could be reproduced by model [2] using $E_{1∕2}^{\mathrm{A}}=-130\mathrm{mV}$ and $E_{1∕2}^{{\mathrm{A}}^{\prime }}=180\mathrm{mV}$ (Figure S7). The observed reduction $E_{1∕2}^{\mathrm{TauD}}$ are within the effective potential range of MG and TA, as is evident for Mb,³¹ and a negligible hysteresis is predicted for [1] with E_1/2 = −130 mV (Figure 4). This observation is important for two reasons. First, if TauD has a single E_1/2 that falls within the effective range of TA and MG, both the reduction and oxidation should occur at this potential unless there is a major difference in the interaction of TauD with the reduced or oxidized forms of both MG and TA, which falls into the definition of [2]. Second, the E_1/2 of Mb (−157 mV by NPSV, no hysteresis) is similar to the observed reduction potential of Fe-TauD (−154 mV, large hysteresis).³¹ The differences in the bimolecular rates of electron transfer between the mediators and Mb or TauD are negligible for the NPSV pulse duration and an equilibrium is expected to be reached before each spectral acquisition.

The observed oxidation $E_{1∕2}^{\mathrm{TauD}}$ may be biased positive if it falls below the effective range for the current FCN concentration and well above the E_1/2 of MG. Such bias can be reduced or eliminated at higher FCN concentrations in both models in agreement with the observed ${\varphi }_{\mathrm{Ox}}^{\mathrm{TauD}}$ (Figure 6); however, the two models predicted an increasingly divergent response of the apparent reduction potential with increasing FCN. The observed ${\varphi }_{\mathrm{Rd}}^{\mathrm{TauD}}$ showed distinctly better agreement with model [2] than with [1].

The complete removal of FCN as a mediator resulted in several characteristic features in the calculated ${\varphi }_{\mathrm{Rd}}^{\mathrm{TauD}}$ and ${\varphi }_{\mathrm{Ox}}^{\mathrm{TauD}}$ profiles that arise from the imbalance between the reduction and oxidation rates³¹ and could be used to further discriminate between models [1] and [2] (Figure 8, Figure S8). The redox transition using [1] starts with an escalating amplitude that follows an n = 1 profile until most of the analyte is reduced and cannot return to the oxidized state due to the kinetic limitations. The ability of mediators to support the reverse transition is limited by mass transfer and/or the bimolecular reaction and, therefore, is not accelerated at increasing E_a. In contrast, model [2] always provides a pathway to return the sample to the oxidized state. In the absence of FCN this oxidation process is accomplished via isomerization of the protein into the low potential form, which can be effectively oxidized by MG and TA. The direct and isomerization pathways are mixed using [2], yielding a broad NPSV profile with an apparent n > 1. The relative contributions of the two pathways depend on E_a so that the overall NPSV amplitude continuously increases with increasing E_a in contrast to a distinctive pseudoequilibrium region using [1]. The low overall amplitudes of the ${\varphi }_{\mathrm{Rd}}^{\mathrm{TauD}}$ and ${\varphi }_{\mathrm{Ox}}^{\mathrm{TauD}}$ profiles in the absence of FCN provide the final argument in support of model [2]. Model [1] predicts a contradiction between $E_{1∕2}^{\mathrm{TauD}}>80\mathrm{mV}$, required to reproduce the ${\varphi }_{\mathrm{Rd}}^{\mathrm{TauD}}$ amplitudes without FCN (Figure S8), and the $E_{1∕2}^{\mathrm{TauD}}\approx 60\mathrm{mV}$, required to reproduce an artificial hysteresis with FCN (Figures 4 and S6). Considering all the experimental conditions examined here, we conclude that the observed redox hysteresis arises mostly from an intrinsic redox-linked isomerization of TauD per model [2], although kinetic limitations may contribute to the observed NPSV response.

Some redox-linked reorganization of the active site is expected due to electrostatic and electronic effects on the coordinated ligands. Distinct structural conformations of metal ligands in reduced and oxidized states lead to changes in E_1/2.^38-43 The magnitude of such changes in TauD greatly exceeds those of other currently known examples with values of <100 mV.⁴⁴ The hysteresis of ≈300 mV (Table S1) is equivalent to a reversible reorganization with ΔG > 7 kcal/mol (n = 1) or pK_a > 5, if coupled to a single protonation event (n_H⁺ = n_e⁻). This observation raises an intriguing question of the extent to which ligand conformation contributes to the unusually low pK_a of the ferric (hydr)oxo F3 species of TauD, detected by transient Raman spectroscopy.^3,8 An uncompensated change in the E_1/2 of 7 kcal/mol would represent a substantial stabilization of TauD following the redox transition. Considering the magnitude of this electrochemical relaxation and the likelihood of redox-linked protonations, it is possible that the observed decrease in $\Delta G_{{\mathrm{e}}^{-}}^{\mathrm{o}}$ is partially compensated by an increase in $\Delta G_{{\mathrm{H}}^{+}}^{\mathrm{o}}$, resulting in a small $\Delta G_{\mathrm{tot}}^{\mathrm{o}}$ in the context of the Bordwell relationship (Figure 1). This possibility can be examined experimentally from the pH dependence of the E_1/2 and associated vibrational changes, which is currently under investigation. The E_1/2 of the ferrous isomer of TauD reported here is over 0.5 V higher than the Fe^2+/3+ transition in CYP450.⁴⁵ If this isomer conformation is transiently retained during the catalytic cycle, the F3/F4 transition in TauD may also have a E_1/2 much higher than that of Cmp I/II of CYP450 while the pK_a of F3 is lower than that of Cmp II. In this case, the ferrous isomer conformation would promote deprotonation of F3 that favors the alkoxide pathway in TauD over the hydroxyl radical rebinding pathway, as found in CYP450. The structural rearrangements may lead to a reduction in the oxidation potential of F4 as a protection mechanism against long-range oxidation if the reaction occurs in the absence of taurine.

The magnitude of the redox hysteresis in TauD suggests that the redox-linked reorganization involves at least one of the metal ligands: His99, Asp101, and His255 (Figure 9a). The redox-difference vibrational changes of the amide mode (Figure 2) suggest that structural changes propagate beyond the first coordination shell. The redox-linked changes in protonatable residues, particularly deprotonation of histidine(s) upon oxidation, can maintain charge neutrality in the active site (a redox Bohr effect) and alter the hydrogen bonding network.⁴⁶ Initial deprotonations in TauD are likely to occur much faster than can be probed by NPSV due to accessibility of the active site for water molecules.⁴⁷ A 2.6 Å distance between N1 of His99 and the peptide carbonyl oxygen of Asn97 suggest fairly strong interactions in Fe(II)-TauD.

Figure 9.

TauD active site structure and possible structural rearrangement. (a) Selected residues at the TauD active site. The carbon atoms of 2OG and taurine are shown in orange. The peptide segment proposed to be linked to structural rearrangement is highlighted by use of stick mode. Selected hydrogen bonding interactions (yellow) and water molecules (red) are shown.⁴⁷ (b) Proposed reversible redox-linked structural rearrangement.

Deprotonation of His99 upon oxidation would disrupt this interaction, allowing the Asn97–Asp101 peptide backbone to undergo further changes. This perturbation could disrupt weak interactions between Asp101 and Trp248 (3.6 Å) in Fe(II)-TauD, allowing bidentate carboxylate binding of Asp101 upon oxidation (Figure 9b), which has been observed in other non-heme iron enzymes.⁴⁸ Such a change involving Asp101 is likely to cause a larger backbone reorganization than the initial deprotonation of histidine residue(s), but it may not be the primary trigger since Asp101 is already deprotonated in Fe(II)-TauD. The interaction of His255 with water 521 also may be involved. Changes in both the carboxyl and imidazole moieties are consistent with redox-difference NPSV spectra of TauD. Vibrational effects of cosubstrate binding suggest that neither metal-bound water molecules nor 2OG directly control isomerization, although both are affected by it. A detailed analysis of the structural origin for the redox-linked switching and its role in catalysis is currently under investigation.

CONCLUSIONS

Our results reveal extensive, fully reversible redox-linked conformational changes in three forms of TauD. The hysteresis between the oxidation and reduction Nernstian NPSV profiles arises primarily from isomerization between two separate ferric/ferrous redox couples of the protein. Quantitative kinetic modeling shows that a redox-linked conformational switching process substantially improves the fitness of simulations compared to the experimental data across various conditions over a simple model with a single redox potential. Changes in the redox potential of up to 0.3 V are attributed to the reversible reorganization of the metal center primary ligands, leading to further reorganization of the protein backbone.