Insight into the mechanism of an iron dioxygenase by resolution of steps following the Fe^IV═O species

Abstract

Iron oxygenases generate elusive transient oxygen species to catalyze substrate oxygenation in a wide range of metabolic processes. Here we resolve the reaction sequence and structures of such intermediates for the archetypal non-heme Fe^II and α-ketoglutarate-dependent dioxygenase TauD. Time-resolved Raman spectra of the initial species with ¹⁶O¹⁸O oxygen unequivocally establish the Fe^IV═O structure. ¹H/²H substitution reveals direct interaction between the oxo group and the C1 proton of substrate taurine. Two new transient species were resolved following Fe^IV═O; one is assigned to the ν_FeO mode of an Fe^III─O(H) species, and a second is likely to arise from the vibration of a metal-coordinated oxygenated product, such as Fe^II─O─C₁ or Fe^II─OOCR. These results provide direct insight into the mechanism of substrate oxygenation and suggest an alternative to the hydroxyl radical rebinding paradigm.

Interest in the mechanism of iron oxygenases arises from their roles in an array of critical biological functions ranging from bacterial biodegradation of xenobiotics and recalcitrant compounds to human drug metabolism and cellular regulation. Many heme and non-heme iron oxygenases are believed to share highly oxidized iron-oxo species as central elements in their reaction mechanisms (1, 2). Several transient oxygen intermediates have been studied extensively in heme enzymes, including compound I-type species of cytochrome P450s (3), but the intermediates occurring during substrate oxygenation have not been directly observed. Even less is known about the transient species in the non-heme Fe oxygenases. An Fe^IV-oxo intermediate was observed in the Fe^II and α-ketoglutarate-dependent dioxygenase TauD (4–6), the archetype of this enzyme family (7), and in the related prolyl 4-hydroxylase (8) or the chlorinating enzymes CytC3 and SyrB2 (9, 10). The Fe-oxygen vibration in TauD at 821 cm^-1 was assigned to an Fe^IV═O stretching mode, whereas an additional oxygen vibration at 583 cm^-1 was not assigned (6). Here, through the use of substrate and media isotopes and varying reaction times, we resolve vibrations associated with three distinct transient oxygen-containing species during TauD catalysis that lead us to propose an alternative to the hydroxyl radical rebinding step that is central to the traditionally accepted mechanism (Fig. 1).

Fig. 1.

Postulated mechanism of taurine oxygenation by TauD. States A, B, C, and F have been experimentally observed; state D is consistent with the chemistry of biomimetic model compounds; the radical rebinding mechanism of state G is inferred from studies of heme-containing oxygenases.

Results

Transient Oxygen Species Detected by Difference Raman Spectroscopy.

The time dependence of the TauD reaction with ¹H- and ²H- taurine was examined by cryogenic continuous-flow Raman spectroscopy (Fig. 2 and Fig. S1). The previously reported oxygen vibrations (6) can be seen as shifts at 825/788 and 578/555 cm^-1 between the ¹⁶O and ¹⁸O derivatives (for ¹H-taurine, Fig. 2A). Intensities of the isotopic shifts diminished rapidly at longer delay times, although the decay of both species was significantly slower and overall intensities were greater with ²H-taurine.

Fig. 2.

Resolution of transient oxygen species of TauD. (A) Raman spectra with ¹H- and ²H-taurine at -36 °C are shown as isotope differences (Gray), which reveal changes in oxygen vibrations as pairs of inverted bands (Fig. S2). Black traces show simulated spectra. Frequencies of a laser plasma line (▴) and a major ethylene glycol peak (♦) are indicated. Previously reported data for ¹H-taurine at 0.22 s (6) are shown for comparison. (B) Intensity profiles of individual F4, F3, and FX species in simulated difference spectra. The ∼800 cm^-1 shift was modeled as (i) separate but overlapped species, F4 and FX (Solid Line), or (ii) as F4 alone (Dashed Line).

The ¹⁶O/¹⁸O difference spectra around 800 cm^-1 at the shortest delay times (0.8 s for ²H- and 0.4 s for ¹H-taurine) could be satisfactorily described by a single ¹H/²H-sensitive isotopic shift at 825/788 and 828/791 cm^-1, although the features were somewhat asymmetrical with both substrates. For ²H-taurine the ¹⁸O mode exhibited a shoulder at the lower frequency side and lower intensity than the better-defined ¹⁶O mode. For ¹H-taurine both oxygen isotope modes were of similar intensity and width, but a shoulder was observed around 815 cm^-1. As the intensity of the ¹⁶O mode decreased with time, a pronounced downshift of the positive bands from 825 and 828 to ∼815 cm^-1 was observed with both substrates. The band (negative) did not show a frequency shift over time. We interpret the clear shift of the ¹⁶O mode as arising from a third oxygen isotope-sensitive species with a ¹⁶O frequency of ∼815 cm^-1 and an isotopic shift of ∼25–30 cm^-1. The three oxygen isotope-sensitive vibrations with ¹⁶O modes at ∼825, 578, and ∼815 cm^-1 are termed here the F4, F3, and FX species, respectively.

Modeling of the Transient Spectra.

Temporal intensity profiles and accurate frequencies of the F4, F3, and FX species observed for ¹H- and ²H-taurine were obtained by spectral global regression analysis (Fig. 2B). Whereas individual early spectra in the 800 cm^-1 region were well described as a single F4 species, global simulation of the complete dataset (Fig. 2B, dashed line) yielded a low fidelity of fit because of a shift from 826 to 815 cm^-1 over time (Fig. S3). To model the observed frequency changes it was necessary to use two independent but overlapping species F4 plus FX (solid lines in Fig. 2B; Fig. S3). Difference spectra at 0.22 s for ¹H-taurine and 0.75 s for ²H-taurine provided a good estimate for frequencies and width of F4. For the dual-shift model, the spectrum of the ²H-taurine derivative at 3 s provided an estimate for the FX species. These values were used as initial guesses in the subsequent global regression analysis of the dual-shift model. Comparison of the regression results for models of F4 only or overlapping F4 and FX shifts are provided in SI Text and Fig. S3. The ¹⁶O frequency of the FX species was 815 cm^-1 (Fig. 2A) with a isotopic shift of ∼29 cm^-1. Unlike the shift of the ¹⁶O mode, the broadening of the ¹⁸O mode with ²H-taurine showed no clear time dependence and was modeled as a vibrational energy transfer common in proteins (11, 12), as described in SI Text.

In contrast to the 800 cm^-1 region, the spectra in the ∼600 cm^-1 region were modeled as a single F3 species. Individual (Fig. 3) and time-dependent (Fig. 2) spectra of the 578 cm^-1 species with ¹H- and ²H-taurine isotopes were satisfactorily modeled by using a single, symmetrical frequency shift with a half-width of 16 cm^-1. This mode in the ¹H-taurine data from the earlier study (0.22 s) was not included in the analysis because of its broadening due to the lower resolution (see SI Text).

Fig. 3.

Sensitivity of TauD intermediates to ¹H/²H substitution. Data are shown as ¹⁶O - ¹⁸O isotopic differences observed at 1.5 s (Gray) along with simulated spectra (Black). Individual contributions of F4 (Red), FX (Blue), and baselines (Dashed Line) are shown. Intensities of individual spectra in both regions were adjusted for clarity. Experimental conditions are the same as in Fig. 2.

The temporal dependencies of the intensities for F4, F3, and FX obtained during global regression of the complete dataset are compared in Fig. 2B. We conclude that all three profiles are noncoincident, with F4 exhibiting the fastest, exponential decay. F3 was delayed in comparison to F4, particularly with ²H-taurine. FX showed a biphasic profile, developing early in the reaction with ²H-taurine, and exhibited the longest lifetime among the three species. ²H-taurine led to an expected increase in the lifetime and intensity of the F4 species (13) and, importantly, had similar effects on the F3 and FX species.

Isotopic Sensitivity.

Visual comparison of transient spectra showed a 3–4 cm^-1 increase in the frequency of F4 from ∼825 to 828 cm^-1 upon ¹H/²H-taurine substitution. Global regression to an F4-only model yielded shifts of +5.1 ± 1.1 (¹H/²H-taurine) and -33.1 ± 5.2 cm^-1 () for the 820.4 ± 1.0 cm^-1 mode. The small amplitude and large error in the shift were caused by the downshift of the ¹⁶O mode at later times. Corresponding shifts in the dual-species F4/FX model confirmed the ¹H/²H sensitivity of F4 but provided more stringent values of +1.8 ± 1.8 and -37.0 ± 0.7 cm^-1 for the 824.6 ± 1.8 cm^-1 mode of F4, respectively (Fig. 3). A part of the observed shift is clearly seen in Fig. 2, and the increased uncertainty in the position and the ¹H/²H sensitivity of F4 are attributed to the spectral overlap with FX. The F3 species showed -23.2 ± 0.6 () and +1.4 ± 0.7 cm^-1 (¹H/²H-taurine) shifts of the 578.3 ± 0.7 cm^-1 mode. Corresponding shifts of the 815.5 ± 1.8 cm^-1 mode of FX were -28.5 ± 1.8 and -2.9 ± 1.7 cm^-1. Media ¹H/²H substitution for the ²H-taurine sample caused additional shifts of -1.0 ± 1.1, -0.4 ± 0.4, and -1.4 ± 0.8 cm^-1 for F4, F3, and FX, respectively, which are insignificant considering the indirect frequency calibration used for ²H media because of solvent frequency shifts.

Mixed Oxygen Isotope Studies.

To ascertain the number of oxygen atoms associated with each of the three transient oxygen species, cryogenic continuous-flow Raman studies were carried out by using the mixed oxygen isotope ¹⁶O¹⁸O (see description in SI Text). The and difference spectra yielded frequencies identical to each other and to those observed for the difference (Fig. 4). Normalized intensities of the isotopic shifts observed with ¹⁶O¹⁸O were half of those observed for the difference. The difference between the ¹⁶O¹⁸O derivative and the average of the and derivatives yielded no new isotopic shifts. The feature-rich double difference spectra characteristic of dioxygen compounds (Fig. S2, Bottom, Right) were not observed for any of the TauD species described in this study, ruling out the presence of an intact O─O bond in any of these intermediates. The featureless difference spectrum shows that all intermediates contain a single atom derived from O₂.

Fig. 4.

Sensitivity of TauD intermediates to ¹⁶O¹⁸O substitution. Intensities of isotopic difference spectra (Gray) between symmetrically ( or ) and asymmetrically (¹⁶O¹⁸O) labeled derivatives were normalized by using an internal standard, and the simulated spectra (Black) were obtained by using the results in Fig. 2. Experimental conditions are the same as in Fig. 3 except for an increased spectral resolution.

Oxygen-Labeled Model Compounds.

To assist in the structural identification of the unique oxygen-sensitive species, we examined the magnitudes of ¹⁶O/¹⁸O isotopic shifts for compounds that resemble potential oxygenation products—isopropanol, propionic acid, and their salts—as shown in Fig. 5 and Figs. S4 and S5.

Fig. 5.

Effect of metal binding on oxygen vibrations of secondary alcohol derivatives. Infrared absorption spectra of (a) isopropanol, (b) Na⁺ isopropoxide, (c) Zn²⁺ diisopropoxide, and (d) Zn²⁺ monoisopropoxide complexes are shown as ¹⁶O - ¹⁸O isotopic difference. Assignment of O─C─C symmetrical and asymmetrical stretching modes is shown by arrows. The ν_OCCsm region in traces b and c is magnified for clarity.

Although the O─C─C symmetrical stretching mode of isopropanol (ν_OCCsm = 818/807 cm^-1, Fig. 5) is very close in frequency to FX (816 cm^-1), the IR and Raman isotopic difference spectra showed a much smaller ¹⁶O/¹⁸O isotopic shift (Δν = 11 cm^-1, trace a). Formation of the Na⁺ isopropoxide complex caused significant changes in oxygen-sensitive vibrations in the ν_OCCsm and ν_OCCas regions where multiple narrow shifts were observed (trace b). When Na⁺ was replaced with a limited amount of Zn²⁺ to form a highly insoluble diisopropoxide complex, vibrational changes were partially reversed and the observed ¹⁶O - ¹⁸O difference spectrum (trace c) was closer to that of isopropanol (trace a) except for the appearance of two narrow shifts in the ν_OCCsm region. Binding of additional Zn²⁺ to form a THF-soluble monoisopropoxide complex resulted in a further downshift of oxygen-sensitive vibrations, a change in their relative intensities, and a significantly larger Δν = 19 cm^-1 of a single mode in the ν_OCCsm region (trace d). This large shift was observed in the absolute spectra of Zn²⁺ monoisopropoxide derivatives (Fig. S4, trace b) excluding the possibility that it is the result of two small overlapping shifts such as observed in the case of Zn²⁺ diisopropoxide derivatives (Fig. S4, trace d).

To identify the ¹⁶O/¹⁸O shift of the OCO scissoring mode (δ_OCO) of singly labeled propionic acid, a 50% ¹⁸O-enriched sample was prepared that is expected to yield a 1∶2∶1 ratio of unlabeled, singly, and doubly labeled isotopes as opposed to mostly doubly labeled isotope in a 90% ¹⁸O-enriched sample. A broad Raman vibration was observed at 844 and 827 cm^-1 for unlabeled and 90% ¹⁸O-enriched propionic acid, respectively; vibrations were not resolved for the 50% enriched sample. Three different well-resolved vibrations were observed at 867, 862, and 850 cm^-1 in the solid state Raman spectrum of the 50% enriched Na propionate (Fig. S5) with relative intensities of 1∶2∶1 corresponding to 0/2, 1/2, and 2/2 exchanged oxygen atoms. The same pattern and isotopic shifts were observed for Zn²⁺ propionate at 906, 894, and 881 cm^-1 (Fig. S5). The ¹⁶O¹⁸O mode observed for both salts showed no evidence of broadening or splitting indicative of asymmetry in binding for both propionate salts.

Discussion

Our cryogenic continuous-flow Raman studies have allowed us to observe three oxygen isotope-sensitive modes with noncoincident temporal profiles. The distinct timing for each feature demonstrates that no two modes arise from the same species (e.g., stretching and bending modes) and the different species are not in rapid equilibrium, such as observed for intermediates in CytC3 and SyrB2 (9, 10). Thus, we conclude that three separate detectable species exist in this reaction, which progress in the order of F4 → F3 → FX (Fig 2B).

Our earlier assignment of ν_FeIV═O in F4 was based on the frequency and amplitude of the isotope shift (6); however, this mode falls close to typical values of the ν_O─O of iron-peroxo complexes. The 578 cm^-1 mode is similarly close to typical values for the ν_Fe─O and δ_Fe─O─O modes of peroxo and superoxo complexes. We utilized asymmetrically labeled ¹⁶O¹⁸O oxygen (Fig. 4) to distinguish unequivocally the number of oxygen atoms associated with F4 and the other two species (14). We observed no new frequencies in the isotopic difference or double difference spectra involving the ¹⁶O¹⁸O derivative of TauD and conclude that a single oxygen atom is present in all three species.

The frequency, ¹⁶O/¹⁸O isotope sensitivity, and ¹H/²H-taurine kinetics of F4 agree with the Fe^IV═O intermediate previously examined by Mössbauer effect, x-ray absorption, and stopped-flow UV-visible spectroscopies (4, 5, 13). ²H-taurine labeling caused a 2 cm^-1 increase in the frequency of F4 (Fig. 3). Solvent ¹H/²H substitution had no discernible effect (< 2 cm^-1) on the observed frequencies (Fig. 3), indicating that no exchangeable protons interact directly with the oxygen atom in F4. Whereas a longer lifetime for the Fe^IV═O species with ²H-taurine was expected, the effect of deuteration on the ν_Fe═O mode provides direct evidence for interactions between the oxo group and substrate hydrogen (15–18).

A single oxygen atom associated with the 578 cm^-1 mode of F3 suggests an Fe─O(H) structure, but F3 showed no sensitivity to substrate or bulk solvent ¹H/²H substitution (Fig. 3). This mode is consistent with the calculated ν_Fe─O of tripodal Fe^III─O^- (19) but is lower than the ν_Fe─O (671 cm^-1) of the only reported non-heme Fe^III─O^- model complex (20). This frequency is at the high end of calculated ν_Fe─OH values (19) and distinct from those of most experimentally measured hydroxyl modes, but it is very close to the ν_Fe(IV)─OH mode of protonated compound II of chloroperoxidase (CPO) at 565 cm^-1 (21) and the ν_Fe(III)─OH modes of di-iron hydroxomethemerythrin at 565 cm^-1 (22) and a synthetic Fe^III─OH model at 574 cm^-1 (23). The ¹⁶O/¹⁸O shifts are similar between all of these species and F3. Unlike F3, however, CPO and hydroxomethemerythrin exhibit characteristic ¹H/²H downshifts of 13 and 5 cm^-1, respectively. Such downshifts are well documented for the Fe─OH₍₂₎ stretching mode of ferric aqua- and hydroxy-hemes (typically 5–20 cm^-1). A few heme ferric-hydroxy complexes lack ¹H/²H sensitivity or even exhibit an upshift with deuteration. This unusually small sensitivity of the ν_Fe─OH mode to deuteration was rationalized by (i) a highly bent geometry of the hydroxyl ligand and (ii) strong hydrogen bonding that leads to an upshift of ν_Fe─O upon deuteration (16–18). Indeed, the crystal structure of the model Fe^III─OH revealed a highly bent hydroxide (∠Fe─O─H = 93°) because of interactions with the iron ligand carboxylate, as well as strong hydrogen bonding between the oxygen and three other caging ligands (23). One could argue that a similar configuration may take place in F3, particularly if iron-bound D101 could rotate to form a hydrogen bond with the hydroxo group. On the other hand, the insensitivity of F3 to bulk media deuteration clearly argues against additional hydrogen bonding to the hydroxo oxygen, without which hydrogen bonding with the carboxylate would weaken. Whereas small rearrangements relative to the static crystal structure could take place, there is no evidence for other steric interference in the active site that could significantly distort the Fe─O─H angle. Although the protonation state of the oxygen in F3 cannot be assigned unambiguously, the structure of the active site and the lack of detectable media ¹H/²H shifts raise a significant probability that the oxo group in F3 is deprotonated.

Although the intensity of FX is low under the present conditions, it is sufficient to discern clear time-dependent spectral changes. Accounting for FX in spectral modeling resulted in a noticeably better fit to the experimental spectra (Fig. S3), and the Δν_16O/18O = -37 cm^-1 observed for F4 in the F4/FX model is a better match for the expected Δν = -36 cm^-1 than the Δν = -33 cm^-1 observed for the F4-only model. Our data indicate a small (∼3 cm^-1) decrease in the observed frequency of FX with ²H-taurine to the extent that its low intensity with ¹H-taurine allows. The slower decay of FX with ²H-taurine implies the kinetic involvement of a substrate proton(s) in the subsequent step. FX must contain a single oxygen atom from O₂, and yet it cannot be assigned to known Fe-oxygen compounds: The 29 cm^{-1 16}O/¹⁸O shift is smaller than the Δν_Fe(IV)═O =  ∼ 36 cm^-1, and the 815 cm^-1 frequency is too high for an Fe─OX stretching mode (ν_typ. < 600 cm^-1) (6), suggesting that FX originates from an oxygenated product of the reaction.

The δ_OCO mode of succinate, which incorporates one atom from O₂ (Fig. 6A) (24), is not likely to give rise to FX because the Δδ =  ∼ 13 cm^-1 of the symmetrical salt of the ¹⁶OC¹⁸O carboxylate is too small compared to that of FX (Fig. S5). This shift may increase in a monodentate Fe-carboxylate complex similar to Zn²⁺ alkoxide (Fig. 5), but the intensity profile of FX with ²H-taurine (Fig. 2B) further argues against its assignment to succinate, which should be formed concurrently with F4. The second atom from O₂ forms the product aldehyde (Figs. 1 and 6) via a transient C₁ alcohol or alkoxide of taurine. An aldehyde would not coordinate to Fe^II, and its oxygen vibrations appear at significantly higher frequencies. The ν_OCCsm of isopropanol (Fig. 5), the simplest secondary alcohol, is very close to FX in frequency but has a much smaller ¹⁶O/¹⁸O shift (Δν = 11 cm^-1). This isotopic shift almost doubled in one of the Zn²⁺-isopropoxide complexes to Δν = 19 cm^-1 (Fig. 5). Whereas it is still smaller than the shift observed for FX, our results show that the geometry of the alkoxide complex had greater effect on oxygen vibrations than the nature of the metal. Thus, it is vibrationally feasible that the formation of a transient Fe^II-alkoxide complex can give rise to the transient FX species. The amplitude of the isotopic shift in TauD can be further affected by the geometry of the complex, the mass and electronic effect of a sulfur atom, and the vibrational interactions at the active site.

Fig. 6.

Possible mechanisms of taurine oxygenation by TauD. The structure of the active site following O─O bond cleavage (A) is the Fe^IV═O species. According to the standard pathway, species A oxidizes substrate to form the Fe─OH complex (B), which is followed by radical rebinding to form an alcohol (C) and decomposition into products (D). Alternative pathways involve formation of an alkoxide (C^′) upon concerted (B → C^′) or stepwise (B → B^′ → C^′ deprotonation via a transient Fe^III─O^- species. Atoms originating from O₂ are grayed; α-ketoglutarate (R₁) and taurine (R₂) are partially abbreviated.

Resonance enhancement of Raman spectra requires electronic absorption near the excitation wavelength (364 nm). Stopped-flow absorption spectroscopy of the TauD reaction provided no evidence of multiple distinctive optical species in the near-UV region at higher temperatures (4, 25, 26), nor were such species seen during in situ optical measurements under cryogenic continuous-flow conditions (6). Thus, either the absorption changes at 300–400 nm are contributed by several unresolved species with broad absorption spectra or the new species do not absorb in this region. Enhancement of F4 may occur via pre- or near-resonance with its 318-nm absorption (4), whereas excitation of FX may be even less favorable implying either a larger energy difference from the electronic transition or a weaker chromophore. Although the Fe^II alkoxo complex should not exhibit charge transfer transitions similar to the Fe^II-αKG-TauD complex (27), the low symmetry of the Fe^II site in TauD may favor spin-forbidden mid-UV transitions below 300 nm, thus allowing for a weak enhancement of oxygen vibrations. In any case, the rapid isotope-dependent disappearance of FX shows that it originates from a transient substrate-derived species.

Possible oxygenation mechanisms following the Fe^IV═O species in the reaction of TauD are depicted in Fig. 6. Binding of α-ketoglutarate and taurine at the active site allows for O₂ binding, O─O bond cleavage, and concomitant formation of CO₂, succinate, and the highly oxidized Fe^IV═O species (F4) (Fig. 1A–F). The oxo group of F4 is poised to oxidize the substrate (Fig. 6) via direct interactions with a proton on the C₁ carbon as revealed by its substrate isotope sensitivity, similar to heme Fe^IV═O species (16–18). A short Fe─C₁ distance in TauD agrees with crystallographic data of the enzyme-substrate complex (28, 29) and with electron spin-echo envelop modulation spectroscopy results that place a substrate C₁ deuteron 1.8 Å from the Fe^II─NO center (30).

According to the generally described mechanism for this enzyme family (7, 31, 32), taurine oxidation proceeds via direct hydrogen atom transfer from C₁ of the substrate to the Fe^IV═O group yielding a Fe^III─OH and a substrate radical (Fig. 6, A → B), supported by a large taurine ¹H/²H kinetic isotope effect on the decay of the Fe^IV species. Our results indicate that a metastable deprotonated Fe^III─O^- species (F3) may be formed as the Fe^IV═O decays. F3 could be formed by rapid proton transfer from Fe^III─OH to a nearby base immediately following hydrogen atom transfer (B → B^′). A less likely direct route (not depicted) may involve an A → B^′ transition via a concerted electron and proton transfer from the C₁ carbon to the Fe^IV═O and a base, respectively. Both pathways will lead to the ¹H/²H isotope-sensitive (33) formation of an Fe^III─O^- species and protonation of a nearby base contingent on the relative pK_as of the hydroxide and the base. In the absence of the base only B is expected.

The classical radical rebinding mechanism applied to TauD posits the transfer of OH^• from Fe^III─OH onto to form the alcohol (B → C), deprotonation, and cleavage of the C₁─S bond with the formation of sulfite, aminoacetaldehyde, and regeneration of Fe^II (C → D). A neutral alcohol will not bind to a metal and is expected to exhibit a much narrower ¹⁶O/¹⁸O shift than we observe; therefore, structure C is not likely to give rise to the FX species, which is vibrationally distinct from an alcohol. We propose that an Fe^II─O─C₁ alkoxy complex of taurine (C^′) is formed instead of an alcohol, tentatively accounting for the observed FX. The formation of C^′ from Fe^III─OH (B → C^′) would require concerted deprotonation, because metal alkoxides undergo rapid hydrolysis. If the oxygen is already deprotonated in F3, formation of FX could occur directly (B^′ → C^′). Both routes would likely be reversible until the C₁─S bond is cleaved (C^′ → D). The substrate proton can subsequently protonate the sulfonate oxygen to maintain charge neutrality, thus explaining the distinctly longer lifetimes observed for F3 and FX with ²H-taurine.

The possible formation of ferric oxide as F3 (B → B^′ or A → B^′) requires a base that is stronger than the Fe^III─OH group. The [Fe^IIIH₃buea(OH)]^- complex has a pK_a of 25 and substitution of Mn^III for Fe^III increased the pK_a of the terminal oxo ligand to 28.3 (34). Oxidation of [Mn^IIIH₃buea(OH)]^- to [Mn^IVH₃buea(OH)]⁰ decreased the pK_a from 28.3 to ∼15 (35). A pK_a of > 15 was also estimated for porphyrinoid [(TBP8Cz)Mn^IV(OH)]⁰ (36) indicating that the macrocycle does not affect the Mn^IV─OH pK_a value. These observations suggest that the Fe^IV─OH in Fe^IVH₃buea(OH)]⁰ or porphyrin analogs will be basic under similar conditions (pK_a≥12). On the other hand, the porphyrin Fe^IV─OH complex in most histidine-ligated peroxidases is significantly more acidic (pK_a < 4) (19). Therefore, heme and non-heme Fe^III─OH complexes in proteins are also likely to be significantly more acidic than in corresponding model compounds. The structure of Fe^II-αKG-TauD suggests that H99 and H255 are neutral and, thus, the Fe^III─OH complex should be neutral (Fig. 6, B^′), similar to most ferric histidine-ligated hemes. In contrast, a net negative charge on [Fe^IIIH₃buea(OH)]^- and cysteine-ligated hemes (37) is likely to contribute to the higher proton affinity of their oxo complexes (19, 38). The paramount role of protein ligands and local charges in determining the pK_a of metal-bound oxygen is further exemplified by siderophores, where the pK_a1 of bound water can shift by as much as 8 pH units (39). Greater variability is expected in the low dielectric environment of a protein.

Direct assessment of the pK_a for an Fe^III─OH is of great significance, especially for iron oxygenases, because the proton affinity of the reduced state has been correlated with the ability of the high valent species to abstract hydrogen in model systems (40), as well as in heme enzymes (19). The deprotonation of Fe^III─OH has not been observed experimentally in any biological process, although Fe^III─O^- states have been generated cryogenically (41). Moreover, the protonation of oxo groups in globins occurs relatively slowly even at T > 160 K in spite of a strong coupling to ¹H, as opposed to peroxidases where protonation occurs at 77 K. Such a difference in proton affinities of Fe^III─O(H) states correlates with the relative basicity of cysteine-ligated Fe^IV─OH in CPO (19, 38) and the “push” of electron density toward oxygen in imidazolate-ligated peroxidases (37) as opposed to structurally similar hemes in globins. The effect of axial ligands illustrates the latitude of ligand-mediated protein control over the reactivity of heme complexes, and an even greater role of metal ligands may be expected in structurally diverse non-heme enzymes.

Conclusions

This study presents the direct observation of three transient iron species occurring during substrate oxygenation by TauD. Resolution of two unique intermediates in addition to characterization of the Fe^IV═O species allowed us to develop an alternative mechanism for catalysis by the Fe^II/αKG-dependent oxygenases. Ferric-oxo and metal-substrate alkoxo complexes similar to those proposed here also may occur transiently in a broader range of Fe oxygenases. Chemical models and molecular dynamics calculations for the observed species will further our understanding of biological oxygen catalysis with implications for a diverse array of related enzymes.

Materials and Methods

Enzyme Purification.

TauD was purified as an apoprotein from Escherichia coli BL21(DE3) containing pME4141 as previously described (42, 43). The isolated enzyme was assayed by reaction of the released sulfite with Ellman’s reagent, providing a specific activity ranging from 2.4 to 3.6 μmol min^-1 mg^-1. Because of the nature of the experiment, large quantities of protein are required; an estimated 17 g of purified TauD was consumed while generating the results reported here.

Isotopic Substitution.

¹H-taurine (Sigma), ²H-taurine (99.5%, C/D/N Isotopes, Inc.), and (≥99%, Isotec Laboratories, Inc.) were used as acquired. Asymmetrically labeled oxygen (¹⁶O¹⁸O) was synthesized as described in SI Text. Bulk solvent ²H substitution of the 50% ethylene glycol plus 25  mM Tris in buffer was carried out by alternating evaporation and dilution with 50% volumes of (99%, Cambridge Isotopes) up to five times to achieve a final exchangeable ²H enrichment > 95%. This substitution resulted in shifting of most of the ethylene glycol Raman vibrations. The pH of the resulting buffer was adjusted while correcting for ²H activity (p²H = 0.4 pH units greater than the pH meter reading). Isotopically substituted model compounds were prepared by oxygen exchange with H₂¹⁸O as described in SI Text.

Sample Preparation.

Anaerobic solutions of 0.5 mM TauD in 25 mM Tris, pH 8.0, 50% (vol/vol) ethylene glycol were prepared by 10 cycles between mild vacuum and Ar gas. Anaerobic solutions of 0.5 mM ferrous ammonium sulfate, 2.0 mM ¹H- or ²H-labeled taurine, 2.0 mM Na α-ketoglutarate, and 0.1 mM Na ascorbate were added to anaerobic TauD and formation of the ternary complex was confirmed optically (43). Sample integrity during Raman measurements was confirmed by recording the visible absorption spectrum before and after the experiment by using a 1-cm flow cuvette and an optical fiber spectrophotometer (model S2000; Ocean Optics).

Cryogenic Continuous-Flow Raman Spectroscopy.

The continuous-flow approach with active mixing used in this study was similar to that implemented earlier (6) but with major improvements of the experimental setup (see SI Text and Fig. S1). Temperature at the coldest point (mixer outlet, -36 ± 1 °C) was controlled by a model 34 temperature controller (Cryogenic Control Systems, Inc.) via a stream of cold N₂ gas and was reproducible to ± 0.1 °C.

During the measurements, TauD samples and oxygenated buffers were continuously mixed in equal volumes, delivered into a rectangular quartz flow cuvette (25 mm long, cross-section 0.25 × 0.8 mm), and probed at ∼5 mm from the mixer outlet with the dead volume of ∼2 μL. Sample syringes were driven by computer-controlled high pressure syringe pump modules (Harvard Apparatus) at constant flow rates of 20–160 μL/ min (after mixing), which resulted in delay times of 0.75–6 s for the current dead volume.

Raman scattering was excited with the 363.8-nm line of an argon ion laser (Coherent, model 70), as described previously (6), with laser power of 60 mW at the point of sampling. The scattered light was collected at 90° geometry by using an F = 0.6 aspherical lens and analyzed by using a single polychromator (model TRIAX 550; Jobin Yvon) equipped with an imaging CCD detector (model 5000; Jobin Yvon). The typical spectral slit width was 8 cm^-1, except for the ¹⁶O¹⁸O and measurements where the spectral slit width was reduced to 5 cm^-1; the spectra were recorded at 0.36 cm^-1/pixel and subsequently smoothed by using a binomial algorithm with the half-width of 2 cm^-1 to preserve experimental resolution. Rayleigh scattering was rejected by using a notch filter (Kaiser Optical Systems). To minimize the effect of velocity distribution in laminar flow along the cuvette walls, the laser beam was focused to 0.15 mm diameter in a 0.25-mm-wide cuvette and only scattering from the central 80% height of the cuvette (0.8 mm) was collected. A two-step calibration of Raman shifts against external and in situ standards was used to increase reproducibility to ± 0.3 cm^-1. For further details on sample excitation and detailed calibration procedures, see SI Text.

Spectral Analysis.

Complete time-resolved datasets of ¹⁶O/¹⁸O difference spectra with ¹H- and ²H-taurine were analyzed simultaneously for frequencies and amplitudes of ¹⁶O/¹⁸O and ¹H/²H-taurine isotopic shifts by using global nonlinear regression analysis using custom routines for Igor Pro (WaveMetrics, Inc.). Reported frequency and amplitude errors represent ± 1 SD of the value obtained during regression analysis. See SI Text for further details.