Spectroscopic and magnetic studies of wild-type and mutant forms of the Fe(II)- and 2-oxoglutarate-dependent decarboxylase ALKBH4 Open Access

The superfamily of Fe(II) and 2OG (2-oxoglutarate, also known as α-ketoglutarate)-dependent dioxygenases (Pfam accession number PF03171) is the largest known non-haem Fe protein family able to carry out hydroxylation reactions of unactivated C–H groups. These enzymes act on a variety of substrates [1,2], and the reaction occurs by reductive activation of molecular oxygen coupled with decarboxylation of the co-substrate 2OG to succinate [3–5]. Within the oxidation process through an Fe(IV)O intermediate, one of the oxygen atoms from O₂ is incorporated into the succinate moiety and the other becomes a hydroxy group in the product [3–5]. The catalytically active site is formed by a mononuclear non-haem Fe centre co-ordinated, in general, by two histidine residues and one carboxylate moiety [6–9]. This site is responsible for the binding of 2OG as well as dioxygen. During turnover, the Fe(II) hexa-co-ordinated states change to a final penta-co-ordinated state, with an open oxygen co-ordination site [6–9], as observed in the three-dimensional structure of the deacetoxycephalosporin C synthase with Fe(II) and 2OG bound (PDB code 1RXG), the first structurally characterized member of this family [10]. However, when uncoupled turnover of 2OG takes place, either in the absence of natural substrates or due to incorrect orientation of substrates in the active site, decomposition of 2OG into succinate and CO₂ may lead to enzyme deactivation.

The Fe(II)/2OG dioxygenase AlkB from Escherichia coli is a repair enzyme, which is induced as part of the adaptive response to alkylation damage. AlkB catalyses demethylation of 1meA (1-methyladenine) and 3meC (3-methylcytosine) in DNA in a reaction where the methyl group is hydroxylated and then released as formaldehyde, thereby regenerating the normal base [11,12]. Additionally, AlkB repairs the structurally analogous lesions 1meG (1-methylguanine) and 3meT (3-methylthymidine) and the bulkier exocyclic etheno and ethano adducts [13–17]. Moreover, AlkB also displays activity on methylated RNA [18]. A bioinformatics analysis revealed eight different mammalian ALKBHs (AlkB homologues), denoted ALKBH1–8 [19], and a more recent study has demonstrated FTO (fat mass and obesity-associated protein) [20] to be a functional ALKBH and thus the ninth member of this family [21]. AlkB-like in vitro repair activities have been reported for ALKBH1 [22], ALKBH2, ALKBH3 [23] and FTO [21], but only ALKBH2 has been convincingly demonstrated to function as a repair enzyme in vivo [24,25].

Recently, it has become clear that the ALKBH proteins are involved in processes other than DNA/RNA repair, as ALKBH8 has been demonstrated to be a tRNA modification enzyme [26–28] and ALKBH1 has been implicated in gene regulation [29]. ALKBH4–ALKBH7 remain completely uncharacterized; in fact, nothing is known about their ability to bind Fe-metal ion, the nature of the so-formed active site(s) and their optical/magnetic fingerprints. From the sequence alignment of putative ALKBH4 orthologues from various organisms, we noted the presence of a cluster of four highly conserved cysteine residues very close to the N-terminus (Figure 1). This cluster is not present in AlkB, and we reasoned that it potentially constitutes a complementary Fe–S binding site. In order to clearly elucidate the nature of the Fe-metal binding by human ALKBH4 in its anticipated binding triad of two histidine residues and one carboxylate motif (His¹⁶⁹-Asp¹⁷¹-His²⁵⁴) (Figure 2) and to address the possible role of the cysteine cluster as an additional Fe–S centre, in the present work we studied the enzymatic activities and the optical and magnetic fingerprints [UV–Vis (UV–visible)/EPR] of the wild-type protein and of two mutant derivatives. One mutant (C15A/C17A) contained an abrogated cysteine cluster, whereas the motif putatively involved in Fe(II) binding was destroyed in the other (H169A/D171A).

The cDNA encoding human ALKBH4 was amplified by PCR and inserted between the NdeI and BamHI sites of the expression vector pET-28a (Novagen). The QuikChange™ Site-Directed Mutagenesis Kit (Stratagene) was used for generation of ALKBH4 mutants. To generate the pET-H169A/D171A plasmid, the primers 5′-CTGCCATTGACCCCGCCCTGGCCGACGCCTGGCTG-3′ and 5′-CAGCCAGGCGTCGGCCAGGGCGGGGTCAATGGCAG-3′ were used, and the primers 5′-CGAAGTCCTTCGGGAAGCCGGTGCCAAGGGCATCCGGACC-3′ and 5′-GGTCCGGATGCCCTTGGCACCGGCTTCCCGAAGGACTTCG-3′ were used for generation of the plasmid pET-C15A/C17A (underlining indicates mutation sites). Plasmids encoding GST (glutathione transferase)-tagged mutants were obtained by subcloning the mutated cDNA fragments into the vector pGEX-6P-2 (GE Healthcare), resulting in the plasmids pGEX-H169A/D171A and pGEX-C15A/C17A respectively (BamHI and EcoRI sites were used for H169A/D171A, and BamHI and SalI sites were used for C15A/C17A). All PCR-generated fragments were verified by DNA sequencing.

Plasmids were transformed into the E. coli strain BL21-CodonPlus(DE3)-RIPL (Stratagene), and protein expression was induced by the addition of 0.075–0.1 mM IPTG (isopropyl β-D-thiogalactopyranoside). Cells were disrupted by French press treatment [two passages through a standard FRENCH® pressure cell (Thermo Electron) at a working pressure of 1000 psi (1 psi=6.9 kPa)], and proteins were purified from clarified lysates using affinity purification. For spectroscopic analysis, ALKBH4 and mutants were expressed as GST-fusion proteins and purified using glutathione–Sepharose 4B medium (GE Healthcare). The GST tag was removed by on-column enzymatic cleavage using PreScission protease (GE Healthcare), and HiTrap Desalting columns (GE Healthcare) were used for buffer exchange (50 mM Hepes, pH 7.0, and 100 mM KCl). His₆-tagged AlkB proteins for decarboxylation assays were purified using TALON® Metal Affinity Resin (Clontech). Proteins were analysed by SDS/4–12% PAGE and subsequent Coomassie Brilliant Blue staining. The metal content (Fe) in the protein preparations was probed by ICP-AEP (inductively coupled plasma atomic emission spectroscopy), with detection at 238.204 nm for the Fe ion emission line.

The reactions were performed as described previously [14]. Briefly, reaction mixtures of 50 μl contained 50 mM Hepes, pH 7.5, 4 mM ascorbic acid, 160 μM 2OG (10% [5–¹⁴C]2OG), 80 μM FeSO₄ (when indicated) and His₆-tagged ALKBH4 (500 pmol of wild-type or mutants) or E. coli His–AlkB (100 pmol). EDTA (1 mM) was added to samples without FeSO₄. Reactions were incubated at 37 °C for 30 min. The amount of [1–¹⁴C]succinate formed was measured by scintillation counting after precipitation of the remaining 2OG with 2,4-DNPH (2,4-dinitrophenylhydrazine).

Concentrated (<300 μM) protein solutions were obtained by centrifugal filter devices (Microcon YM-10) in 50 mM Hepes, pH 7.0, and were then made anaerobic in airtight vessels by several rounds of vacuum treatment and argon exchange using the Schlenk technique [29a]. The minimum amount (≤2 mol per mol of protein) of a freshly prepared anaerobic solution of Na₂S₂O₄ (5.7 mM in 50 mM Hepes, pH 7.0) was then added to the protein solutions using airtight syringes. Stoichiometric amounts of Fe(II) ammonium sulfate were added to the protein samples in a similar way, using anaerobic stocks. In some cases, small amounts (~20 μM) of Methyl Viologen were added in order to monitor that strict anaerobic conditions were maintained within reconstitution procedures. Anaerobic solutions of 2OG and Na₂S in 50 mM Hepes, pH 7.0, were freshly prepared and added to the protein samples at a 10-fold excess. These solutions were transferred to quartz EPR tubes through airtight syringes that were maintained under argon, and then the tubes were sealed with Precision Seal® rubber septa (Sigma–Aldrich). Nitric oxide (NO) gas was added through airtight syringes directly into the quartz EPR tubes containing the sample protein, or to airtight 1 cm optical path quartz cuvettes. Samples were maintained at cold temperatures (≤10 °C). Sample oxidation was obtained by air exposure followed by slow injection of O₂ (~90% purity, ~5 min). In the experiments employing reconstitution in the presence of 2OG/Na₂S/Na₂S₂O₄, sample oxidation was obtained by injection of 50 μl of O₂-saturated 50 mM Hepes, pH 7.0, into EPR tubes containing 100 μl of protein sample, followed by fast quenching in liquid nitrogen. The low-temperature EPR spectra were acquired on a Bruker Elexsys 560 instrument equipped with a dual-band X-resonator (Bruker) and by using a ESR900 He-flow cryostat (Oxford Instruments). UV–Vis spectra were recorded on a HP 8452A diode-array spectrophotometer (Hewlett Packard).

Like several of the Fe(II)/2OG dioxygenases, E. coli AlkB possesses the ability to perform decarboxylation of 2OG to succinate even in the absence of primary substrate [11,12]. Since the primary substrate of ALKBH4 is still unknown, we took advantage of such uncoupled co-substrate conversion to test the in vitro enzymatic activity of ALKBH4. For this purpose, recombinant His₆-tagged ALKBH4 and two mutants were purified from E. coli (Figure 3A) and subsequently assayed for 2OG decarboxylation activity. Indeed, ALKBH4 was able to catalyse succinate formation in the presence of the co-factor Fe(II), although not to the same extent as AlkB (Figure 3B). The decarboxylation activity was reduced to background levels when Fe(II) was not included, demonstrating the requirement for Fe, which is indicative of AlkB proteins. Likewise, the H169A/D171A mutant was devoid of the ability to perform 2OG turnover. The C15A/C17A mutant displayed a decarboxylation activity similar to that of the wild-type ALKBH4 protein, suggesting that the cysteine motif is not involved in Fe(II) co-ordination. Taken together, these results experimentally confirmed that ALKBH4 is a Fe(II)/2OG-dependent decarboxylase.

The electronic/magnetic fingerprints of the native and mutant proteins were then addressed by low-temperature EPR measurements down to cryogenic temperatures. We probed different conditions where the metal co-ordination environment of Fe reconstituted ALKBH4, and mutant proteins were analysed both in the reduced [Fe(II)] and oxidized [Fe(III)] states, as well as in the presence or absence of 2OG/succinate. The results are summarized in Table 1. It should be noted that the isolated proteins (ALKBH4 and mutants), even when kept in iced-cooled solutions (4 °C), aggregated and then precipitated at concentrations higher than 0.5 mM. The process was accelerated greatly when an excess (e.g. 5-fold) of dithionite was added to the solution when anaerobic Fe(II)-reconstitution procedures were carried out. Nevertheless, low protein concentrations (<0.3 mM) and a slight excess of Na₂S₂O₄ (approx. twice greater) allowed efficient reconstitution of the apo-proteins with the Fe(II) metal ion, hence avoiding precipitation. The purified ALKBH4 protein, in the absence of added cofactors, gave rise to an extremely weak high-spin Fe(III) signal at g≈4.3 (Figure 4A). This resonance did not increase upon addition of oxidants (e.g. Na₂IrCl₆). This agrees with the ICP-AES findings that indicated the presence of Fe traces only in the purified protein. Traces of Cu(II) ion as an impurity (around g≈2.1) were also detected, with relative amounts that changed from sample to sample. When ALKBH4 and the C15A/C17A and H169A/D171A mutants were incubated with stoichiometric amounts of Fe(II) under reducing conditions, either in the presence or absence of 2OG/succinate, the recorded low-temperature EPR spectra did not show any Fe-related signal (Figure 4B). It is important to note that the only resonance signal emerging in the spectra was always associated with the presence of MV^•+ (Methyl Viologen radical cation) (g=2.00 in Figure 4B), which was used as an internal indicator to probe the attainment of rigorous anaerobic conditions. These results can be interpreted at X-band EPR with either (i) an integer-spin system having a high-spin Fe(II) state (S=2, where S is the ground-state spin) characterized by a large axial ZFS (zero-field splitting) term called δ in EPR or Δ in MCD (magnetic CD), as it occurs in many mononuclear non-haem Fe proteins where such configuration is usually not spectroscopically accessible, or alternatively (ii) by a low-spin Fe(II) state (S=0) [1,2,30]. NO has been successfully used as a probe for such centres [31], since it can convert the S=2 species into an EPR-active S=3/2 species, as observed, for example, in IPNS (isopenicillin N synthase) [32], in extradiol-cleaving catechol dioxygenases such as 2,3-CTD (catechol 2,3-dioxygenase) and 4,5-PCD (protocatechuate 4,5-dioxygenase) [33,34], as well as in non-haem Fe model complexes [35]. Upon anaerobic addition of NO to Fe(II)-reconstituted ALKBH4 and C15A/C17A protein in the absence of 2OG, a strong axial EPR signal centred at a g_eff of 3.97 developed (g-tensor components at g₁=4.08, g₂=3.97 and g₃=2.00), together with a second asymmetric resonance at g≈2.00 (Figure 4C). The latter signal contained features characteristic of a small amount of NO radical being trapped in the frozen matrix solution. Similarly, when anaerobic reconstitution of the ALKBH4 and C15A/C17A proteins was carried out with both Fe(II) and a 10-fold excess of 2OG, a strong axial resonance signal centred at g_eff=3.97 developed after NO addition, being accompanied by a small highly asymmetric signal around g≈2.00 (Figure 4E). Analogous spurious signals at g≈2.00 are found in other non-haem Fe proteins, such as 4,5-PCD [36] and IPNS [32]. Therefore in both cases a nearly axial S=3/2 Fe(II)–NO adduct formed with the E (rhombic) versus D (axial) term |E/D|≈0.010.

Similarly, anaerobic reconstitution of the proteins with both Fe(II) and a 10-fold excess of the product succinate, followed by addition of NO, led to a single strong resonance signal around g_eff=3.93 that contained a slightly higher degree of rhombicity (|E/D|≈0.015) (Figure 4F). The H169A/D171A protein behaved very differently. After being incubated with Fe(II) under anaerobic conditions, a very strong signal centred at g≈2 appeared immediately after addition of NO. Only a very weak resonance around g≈4 could be detected, with small variations in relative intensities from preparation to preparation. These findings confirm that when the histidine-carboxylate Fe-binding region is abrogated, only a much smaller fraction of Fe productively forms stable S=3/2 Fe–NO complexes (Figure 4D). Upon addition of oxygen, the ALKBH4 protein sample that contained 2OG (Figure 4H), and also the sample that was deficient in co-substrate (Figure 4I), displayed formation of strong similar signals with broad shoulders at g_eff=4.26, being accompanied by the appearance of weak resonances at g_eff=9.40. Those signals originate from transition involving the middle and ground Kramer's doublets respectively and are typically found in high-spin rhombic Fe(III) centres (|E/D|≈0.25). The same resonances developed in the C15A/C17A mutant upon Fe(II) oxidation. Similarly, a strong signal formed at g_eff=4.28 upon oxidation of the Fe(II) H169A/D171A mutant (Figure 4G). However, the presence of broad shoulders like those of the Fe(III) forms of both ALKBH4 and the C15A/C17A mutant were not detected. Such EPR resonance can therefore be interpreted as arising from spurious or non-specifically bound Fe(III) metal ion. Surprisingly, addition of oxygen to the metal-reconstituted ALKBH4 and C15A/C17A proteins in the presence of succinate resulted in protein precipitation. The reason for this is presently unknown.

Motivated by the possibility that the N-terminal cysteine-rich motif might provide an Fe–S cluster, we measured the EPR spectrum of ALKBH4 in the presence of inorganic S and Fe(II). Anaerobic incubation of ALKBH4 with 2OG and Fe(II) in the presence of Na₂S (10-fold excess), followed by addition of O₂ and fast quenching at liquid-N₂ temperature (77 K, –196 °C), resulted in a transient EPR-active species that exhibited resonances at g_eff=8.24, 5.56 and 4.26 (Figure 5). Upon thawing, this new species relaxed fast (~20 s) to the high-spin rhombic Fe(III) ion with |E/D|;≈0.25, showing resonances similar to those described in Figure 4(H). Identical behaviour was observed for the C15A/C17A mutant, but not for the mutant H169A/D171A. These effects suggest that the structural changes occuring in the active site within the Fe(II)→Fe(III) oxidation process are modulated further by the presence of Na₂S. It is important to note that similar signals and a similar decay process have been reported in the non-haem Fe(III) centre present in photosystem II, for example, decrease of the g_eff=8.07 and g_eff=5.58 signals being accompanied by induction (increase) of the signal at the g_eff=4.26 signal, as a result of UV-B irradiation (280–320 nm) [37,38]. These types of resonances have previously been analysed by Weisser et al. [39] in terms of ZFS-distributed S=5/2 systems in Fe–catecholato complexes containing moderate rhombic distortion (|E/D|≈0.12). Furthermore, EPR spectra similar to those reported in Figure 5, although not transient, are observed in Fe(III)–catecholato complexes as model systems of intradiol-cleaving catechol dioxygenases [40], in PAH (phenylalanine hydroxylase) in the presence of catecholamine feedback inhibitors or even when recording PAH in Tris buffer [41,42].

In parallel with the low-temperature EPR experiments, the optical fingerprints of the various forms of ALKBH4 and mutants [Fe(II)/Fe(III) in the presence of, and without, 2OG] were probed by UV–Vis absorption spectroscopy. The electronic spectrum of the purified ALKBH4 protein did not show absorption features in the visible region (Figure 6A, dashed line). Similarly, the C15A/C17A and H169A/D171A proteins gave featureless spectra. Upon reconstitution with Fe(II), but in the absence of 2OG, both ALKBH4 and the C15A/C17A mutant developed a broad and composite absorption band, extending from 400 nm (λ=404 nm, ϵ≈570 M⁻¹·cm⁻¹) to 600 nm (λ=490 nm, ϵ≈260 M⁻¹·cm⁻¹) (Figure 6A, dashed and dotted line). However, the strong absorption band around 320 nm, due to the presence of Na₂S₂O₄ (reducing equivalents) in the medium, masked an additional high-energy absorption signature, around 350 nm (ϵ≈920 M⁻¹·cm⁻¹), as substantiated in those samples reconstituted with Fe(II) without the supply of reducing agents (Figure 6A, continuous line). When a 10-fold excess of 2OG was added under reducing conditions, a slightly different chromophore (λ=430 nm, ϵ≈440 M⁻¹·cm⁻¹) developed quickly in both ALKBH4/Fe(II) and the C15A/C17A/Fe(II) protein (Figure 6B, dotted line), with an absorption envelope similar to that observed by Henshaw et al. [43] in E. coli AlkB in the presence of Fe(II)/2OG. Upon oxidation, both ALKBH4/Fe(II) and the C15A/C17A/Fe(II) mutant showed the formation of an absorption band centred at 530 nm (ϵ≈920 M⁻¹·cm⁻¹), being accompanied by the appearance of a shoulder at higher energy (λ_max=370 nm, ϵ≈2000 M⁻¹·cm⁻¹) (Figure 6A, continuous line). Similarly, upon oxidation of Fe(II)- and 2OG-reconstituted ALKBH4 and C15A/C17A protein, an absorption band that was slightly blue-shifted (λ_max=520 nm, ϵ≈860 M⁻¹·cm⁻¹) was detected together with a shoulder at λ_max=404 nm (ϵ≈1580 M⁻¹·cm⁻¹) (Figure 6B, continuous line). None of the above-reported absorptions were observed in the H169A/D171A mutant, neither in the Fe-reduced nor -oxidized state. Reconstitution of ALKBH4 and the C15A/C17A protein with a 5-fold excess of Fe(II) metal ions prior to oxidation did not result in any further increase of the absorption features in the visible regions. These findings reinforce the hypothesis that (i) ALKBH4 is mononuclear in Fe content, (ii) the cysteine-rich N-terminal site does not provide an additional binding centre for the Fe(II)/Fe(III) metal ion, and furthermore (iii) an intact histidine-carboxylate site is necessary for productive Fe binding, in agreement with the results obtained in the 2OG-decarboxylation-activity assay. Since the optical features of both ALKBH4 and the C15A/C17A protein are clearly modulated by the presence of 2OG, these differences may indeed mirror modifications in the Fe co-ordination environment upon interaction with the substrate, such as the displacement of previously co-ordinated water molecules to the metal Fe. In the present study, the development of rather strong chromophores around 520–530 nm in the Fe(III) ALKBH4 and C15A/C17A protein forms closely resemble the optical spectra observed in synthetic models of non-haem Fe(III)–catecholato/phenolato complexes [40]. Therefore the spectroscopic behaviour displayed especially for the Fe(III) forms of ALKBH4 and the C15A/C17A mutant is different from that previously observed for analogous non-haem Fe proteins, such as AlkB and the Fe(II)/2OG-dependent TauD (taurine dioxygenase). AlkB exhibits an absorption band at λ_max=560 nm (ϵ=1258 M⁻¹·cm⁻¹) [44] or, as found in other reports [12,43], a much weaker broad band at 450–500 nm, whereas TauD shows an absorption band at λ_max=530 nm (ϵ=140 M⁻¹·cm⁻¹) [45], when both 2OG and Fe(II) are present under strict anaerobic conditions. In these proteins, exposure to oxygen in the absence of primary substrate causes the development of greenish-brown chromophores. In TauD (λ_max=550 nm, ϵ=700 M⁻¹·cm⁻¹), the absorption envelope in the oxidized protein arises from a self-hydroxylation reaction of a tyrosine residue (Tyr⁷³) near the non-haem metal site [45]. In AlkB, the formation of a greenish-brown chromophore upon oxidation (λ_max=595 nm, ϵ=960 M⁻¹·cm⁻¹) indicates a similar self-hydroxylation process that involves Trp¹⁷⁸ [43]. In order to further study the O₂-binding properties of ALKBH4, we followed the optical changes of the Fe(II) metal core upon interaction with the dioxygen analogue NO, as performed in the EPR experiments reported earlier. Addition of NO to an anaerobic solution of ALKBH4 containing Fe(II), 2OG and Na₂S₂O₄ produced a strong yellowish-green chromophore, with a broad absorption band around 580 nm (ϵ≈620 M⁻¹·cm⁻¹) being accompanied by a more intense peak at 436 nm (ϵ≈2430 M⁻¹·cm⁻¹) (Figure 6C, continuous line). A similar spectrum was obtained for the C15A/C17A mutant (results not shown). These optical features closely resemble those reported for other non-haem Fe proteins [46] anaerobically treated with NO, such as PAH (λ_max=440 nm, ϵ≈2300 M⁻¹·cm⁻¹) in the presence of its co-substrate tetrahydropterin [47], which exhibits, in addition, similar EPR resonance to our NO complex. However, after injection of oxygen, the Fe(II)–NO adduct decayed into a greenish-brown chromophore (λ=580 nm, ϵ≈620 M⁻¹·cm⁻¹, and λ=770 nm, ϵ≈270 M⁻¹·cm⁻¹) (Figure 6C, dotted line), without recovery of the optical features for the Fe(III) form shown previously (Figure 6B, continuous line). This might indicate that the protein becomes more sensitive towards side reactions upon NO binding and subsequent oxidation, since the recorded envelope in this case is very similar to the one observed in AlkB upon Trp¹⁷⁸ hydroxylation [43]. Large differences can also be seen between ALKBH4 (and the C15A/C17A mutant) and the H169A/D171A mutant with respect to NO reactivity. After addition of NO to the colourless solution of the H169A/D171A/Fe(II)/2OG protein under anaerobic conditions, a yellowish solution formed without the emergence of any of the Fe–NO characteristic absorption bands, apart from a very broad absorption envelope, extending from 350 nm to 700 nm (Figure 6C, dashed and dotted line). This result supports the previous observation from the low-temperature EPR spectra (Figure 4D) that only a minor amount of Fe–NO adduct can be formed in the case of the H169A/D171A mutant, with the majority of the supplied Fe metal ions non-specifically bound to the protein.

In the present work, we have addressed the characteristics (electronic and magnetic) of the Fe metal core in the human AlkB homologue ALKBH4, and we have shown that an intact His¹⁶⁹-Asp¹⁷¹-His²⁵⁴ motif is necessary for productive Fe binding and for decarboxylation activity towards 2OG. Furthermore, we have shown that the N-proximal cysteine-rich motif of ALKBH4 is not involved in Fe binding. Although the biological function of ALKBH4 remains elusive, its ability to bind and decarboxylate 2OG after metal reconstitution, together with the optical and spectroscopic fingerprints of its suggested catalytic centre, confirm for the first time that ALKBH4 truly belongs to the class of 2OG-dependent mononuclear non-haem Fe proteins. Additional work is needed to determine the nature of the transient species detected within the reoxidation process in the presence of Na₂S.

ALKBH

AlkB homologue

FTO

fat mass and obesity-associated protein

GST

glutathione transferase

ICP-AEP

inductively coupled plasma atomic emission spectroscopy

IPNS

isopenicillin N synthase

IPTG

isopropyl β-D-thiogalactopyranoside

MV^•+

Methyl Viologen radical cation

2OG

2-oxoglutarate

PAH

phenylalanine hydroxylase

4,5-PCD

protocatechuate 4,5-dioxygenase

TauD

taurine dioxygenase

UV–Vis

UV–visible

ZFS

zero-field splitting

Linn Bjørnstad planned experiments, produced wild-type and mutant proteins, undertook activity measurement, analysed all of the results and wrote the manuscript. Giorgio Zoppellaro planned experiments, performed EPR and light-absorption experiments, analysed all of the results and wrote the manuscript. Ane Tomter performed EPR experiments, analysed results and wrote the manuscript. Pål Falnes designed wild-type protein and mutants, provided experience in AlkB protein, analysed all of the results and wrote the manuscript. Kristoffer Andersson planned physical chemistry experiments, provided experience in non-haem Fe proteins, analysed all of the results and wrote the manuscript.

This work was supported by the PEOPLE Marie Curie Actions Intra-European Fellowship within the 7th European Community Framework Programme [grant number PIEF-GA-2009-235237 (to G.Z.)], the Norwegian Cancer Society [grant number PR-2007-0132 (to L.G.B.)], the Research Council of Norway [FUGE program grant number 159013/S10 (to P.Ø.F.) and grant number 177661/V30 (to K.K.A.)] and the Polish-Norwegian Research Fund [grant number PNRF-143-AI-1/07 (to P.Ø.F.)].