A peroxodiiron(III) intermediate mediating both N-hydroxylation steps in biosynthesis of the N-nitrosourea pharmacophore of streptozotocin by SznF

Abstract

The alkylating warhead of the antineoplastic drug streptozotocin (SZN) contains an N-nitrosourea moiety constructed from N^ω-methyl-l-arginine (l-NMA) by the multi-domain metalloenzyme SznF. The enzyme’s central heme-oxygenase-like (HO-like) domain sequentially hydroxylates N^δ and N^ω’ of l-NMA. Its C-terminal cupin domain then rearranges the triply modified arginine to N^δ-hydroxy-N^ω-methyl-N^ω-nitroso-l-citrulline, the proposed donor of the functional pharmacophore. Here we show that the HO-like domain of SznF can bind Fe(II) and use it to capture O₂, forming a peroxo-Fe₂(III/III) intermediate. This intermediate has absorption- and Mössbauer-spectroscopic features similar to those of complexes previously trapped in ferritin-like diiron oxidases and oxygenases (FDOs) and, more recently, the HO-like fatty acid oxidase UndA. The SznF peroxo-Fe₂(III/III) complex is an intermediate in both hydroxylation steps, as shown by the concentration-dependent acceleration of its decay upon exposure to either l-NMA or N^δ-hydroxy-N^ω-methyl-l-Arg (L-HMA). The Fe₂(III/III) cluster produced upon decay of the intermediate has a small Mössbauer quadrupole splitting parameter, implying that, unlike the corresponding product states of many FDOs, it lacks an oxo-bridge. The subsequent decomposition of the product cluster to one or more paramagnetic Fe(III) species over several hours explains why SznF was previously purified and crystallographically characterized without its cofactor. Programmed instability of the oxidized form of the cofactor appears to be a unifying characteristic of the emerging superfamily of HO-like diiron oxidases and oxygenases (HDOs).

Graphical Abstract

INTRODUCTION

The enzyme SznF from Streptomyces achromogenes var. streptozoticus NRRL 2697 transforms N^ω-methyl-l-arginine (l-NMA) to N^δ-hydroxy-N^ω’-methyl-N^ω-nitroso-l-citrulline,¹ a proposed precursor to the N-nitrosourea pharmacophore of the pancreatic cancer drug streptozotocin (SZN).^2,3 This overall eight-electron oxidation proceeds in three iron- and dioxygen-dependent steps (Scheme 1) in autonomous active sites harbored within two of the enzyme’s three domains (Figure 1A). The central domain mediates sequential hydroxylations of the unmodified guanidino nitrogen atoms of l-NMA,¹ with the internal nitrogen (N^δ) being oxidized first.^1,4 The C-terminal domain, which has the cupin fold characteristic of a large family of mononuclear nonheme iron enzymes (Figure 1B),^5,6 promotes the last step, in which the triply-modified guanidine moiety of N^δ,N^ω-dihydroxy-N^ω’-methyl-l-Arg (l-DHMA) is rearranged to the N-methyl-N-nitrosourea (Scheme 1).¹ An x-ray crystal structure of the SznF homodimer revealed a mononuclear iron cofactor facially coordinated by three histidine ligands in the cupin domain. The structure did not reveal the nature of the iron cofactor in the central domain, because this site exhibited only minimal metal occupancy in addition to apparent disorder in regions of the protein predicted to provide iron ligands (Figure 1C).¹

Scheme 1.

Sequential N-oxygenation (HDO domain) and oxidative rearrangement (cupin domain) reactions catalyzed by SznF in the biosynthesis of SZN.

Figure 1.

The domain architecture of SznF. (A) Ribbon diagram of the SznF dimer (PDB accession code 6M9R) colored by domain. The enzyme contains an N-terminal structural (dimerization) domain (blue) and two different functional domains (violet, teal). (B) A His₃-coordinated mononuclear iron cofactor in the cupin domain (teal) promotes the final rearrangement of l-DHMA. (C) The central HDO domain (violet) catalyzes two sequential N-hydroxylations of the l-NMA substrate using an iron cofactor that was not observed in the structure. Candidates for iron ligands, selected on the basis of their conservation with ligands seen in structures of CADD and UndA, are highlighted in stick format.

The central domain of SznF has a tertiary structure similar to that of heme oxygenase (HO).⁷ The HO-like architecture has recently been identified in a number of microbial enzymes that appear to constitute an emerging family of O₂-activating nonheme diiron enzymes⁸ (HDOs, for heme-oxygenase-like diiron oxidases/oxygenases). The founding member of this family, UndA, oxidatively decarboxylates medium-chain, linear, saturated fatty acids (n-C_xH_2x+1CO₂H, x = 9 or 11) to the corresponding C_x 1-alkenes.^9,10,11 Although UndA was originally proposed to employ a mono-iron cofactor,⁹ its structural similarity to the functionally unassigned CADD protein (Chlamydia protein associated with death domains),^12,13 which was shown by x-ray crystallography to harbor a dimetal cluster,^10,11 suggested that its iron cofactor might be dinuclear. Indeed, a pair of recent studies established that UndA uses the reduced (II/II) form of a diiron cofactor to activate O₂^10,11 via a peroxo-Fe₂(III/III) complex.¹¹ The HDO domain of SznF conserves all six metal-ligand residues of CADD and UndA.¹ Substitution of any one of these amino acids was shown to abolish both N-hydroxylase activities of the HDO domain, consistent with the inference that they serve as cofactor ligands. Interestingly, for both UndA and SznF, multiple, directed attempts to obtain crystal structures with intact diiron cofactors by co-crystallization or metal soaking resulted in, at best, only fractional occupancy, hinting at a common cofactor instability.^1,9,11

In this study, we show that, despite this apparent deficiency in iron binding, cofactor stability, or both, apo SznF isolated from overproducing Escherichia coli in M9 medium with Mn(II) supplementation readily takes up added Fe(II) and then rapidly captures dioxygen to form a transient, oxidized diiron-O₂ adduct in its HDO domain. This complex, which decays in a few seconds at 5 °C, exhibits an intense visible absorption feature with its peak (λ_max) at ~ 629 nm and a Mössbauer quadrupole doublet with relatively high isomer shift (δ). Both features are similar to those associated with µ-peroxo-Fe₂(III/III) complexes in ferritin-like diiron oxidases and oxygenases (FDOs)^8,14 and the HDO, UndA.¹¹ Whereas both rapid formation and accumulation of the UndA intermediate require that the fatty acid be bound in the active site (termed substrate triggering), formation of the SznF peroxo-Fe₂(III/III) intermediate is not impacted by the presence of substrate. By contrast, both l-NMA and l-HMA (the product of the first N-oxygenation step and substrate of the second) accelerate decay of the peroxo-Fe₂(III/III) complex with hyperbolic dependencies on substrate concentration (i.e., exhibiting saturation). These observations establish that, as was previously reported for the FDO N-oxygenases AurF and CmlI,^{15,16,17,18,19,20} a peroxo-Fe₂(III/III) complex is an intermediate also in the two reactions catalyzed by the SznF HDO domain, either as the hydroxylating complex itself or as a precursor that is induced by substrate binding to convert to the (non-accumulating) hydroxylating intermediate. The development of a new Mössbauer quadrupole doublet with unusually small value of the quadrupole splitting parameter, ΔE_Q, upon decay of the intermediate implies the formation of an Fe₂(III/III) product cluster lacking an oxo bridge. Subsequent spontaneous decomposition of this cluster to uncoupled mono-iron(III) species rationalizes the instability of the cofactor suggested by earlier work.^1,11,12 This property sets HDO N-oxygenases apart from their FDO counterparts, with their stably bound Fe₂(III/III) cofactors.^8,14

RESULTS

Detection of an Intermediate in the HDO Domain of SznF by Stopped-Flow Absorption Spectroscopy (SF-Abs).

Consistent with the previous report,¹ SznF bearing an N-terminal hexahistidine affinity tag and purified by chromatography on Ni(II)-nitrilotriacetate-agarose resin following over-expression in E. coli grown in LB medium had very little iron bound (< 0.02 equiv) and nearly undetectable levels (< 0.01 equiv) of other potentially relevant transition metals (Table S1). Despite its apparent lack of affinity for iron, mixing of an anoxic solution of the protein containing 3 molar equivalents of Fe(II) with O₂-saturated buffer resulted in development of a transient optical absorption feature with maximum absorbance (λ_max) at 629 nm (Figure 2A). At 5 °C, this feature developed to maximum intensity in ~ 0.4 s and decayed with a rate constant of 0.34 ± 0.03 s⁻¹ (Figure 2B, blue). This rapid development and decay suggested that the associated species could be a catalytic intermediate.

Figure 2.

SF-Abs data demonstrating accumulation of an intermediate in the reaction of the Fe(II)•SznF complex with O₂ at 5 °C. (A) Absorption spectra acquired at the indicated reaction times after exposure of the Fe(II)•SznF reactant to O₂. (B) Absorbance at 629 nm (A₆₂₉) as a function of time in the reaction shown in panel A (blue) and in otherwise identical reactions containing 0.50 mM of either l-NMA (black) or l-HMA (purple). (C) A₆₂₉-versus-time traces following exposure of the pre-formed intermediate to l-NMA at the final concentrations indicated. The inset plots the values of the observed rate constant for decay (k₂) at each concentration with a fit of the equation for a hyperbola to the data (solid line). (D) A₆₂₉-versus-time traces from the corresponding reactions with the second substrate, l-HMA. The conditions of each experiment and the procedures used in the regression analysis (solid lines in panels B – D) are provided in the Experimental Procedures.

We considered that Fe(II)-mediated O₂ activation in the absence of substrate during expression and purification could – as seen previously in other systems²¹ – degrade the recombinant SznF and diminish its capacity to take up iron. To test this possibility, we supplemented iron-poor over-expression cultures with Mn(II), with the intent of blocking the presumptive Fe(II) sites and preventing redox cycling. Under identical reaction conditions, the protein obtained by this procedure supported accumulation of more of the absorbing complex (Figure S1A, blue trace) than did protein obtained following over-expression in iron-replete medium (red trace). We used this optimized expression procedure to prepare all of the SznF protein samples used in this study.

Because SznF has two domains, each requiring iron for its activity, we carried out experiments to assess the location of the absorbing complex within the protein. The prior study reported that a variant of SznF, with all three iron ligands in the C-terminal cupin domain (His 407, His 409 and His 448) replaced by non-coordinating alanine residues (SznF-Δcupin), is able to promote both N-oxygenations but not the final rearrangement.¹ We found that this variant protein also supports accumulation of the 629-nm absorbing complex (Figure S1A–B, black traces). Another variant SznF protein lacking Glu 215, the HDO domain residue that aligns with (known or predicted) bridging glutamates in CADD and UndA,^10,11,12 was previously shown not to mediate the N-oxygenation steps.¹ We found the E215A variant of SznF to be incapable of supporting accumulation of the 629-nm-absorbing complex (Figure S1A, gray). These results imply that the transient complex forms in the HDO domain and is likely an intermediate in one or both of the N-oxygenation steps.

Evidence that the Absorbing Intermediate Mediates both N-oxygenation Steps.

A recent study of the HDO, UndA, documented accumulation of a peroxo-Fe₂(III/III) intermediate in its oxidative decarboxylation of lauric acid.¹¹ This work showed that substrate is required for accumulation of the peroxo-Fe₂(III/III) complex. By contrast, we found that inclusion of either l-NMA or l-HMA at 1.0 mM in the Fe(II)/SznF reactant solution (giving 0.5 mM after mixing) did not significantly impact the kinetics of formation of the absorbing complex upon mixing with O₂-saturated buffer (Figure 2B, black and purple traces). The presence of either substrate did, however, markedly accelerate decay of the complex, as implied by the diminished amplitudes and faster disappearance of the transient feature. We observed the same behavior – acceleration of decay but not formation – in the reaction of the Δcupin variant in the presence of l-NMA or l-HMA (Figure S1B). For the case of l-NMA, we used liquid chromatography coupled to mass spectrometry (LC-MS) to confirm that accelerated decay of the intermediate in the presence of the substrate reflects formation of the hydroxylated product. In the chromatograms of samples prepared from complete reactions [containing SznF (Figure 3A) or SznF-Δcupin (Figure 3B), 3 equiv Fe(II), 2 equiv l-NMA, and excess O₂], the peak at m/z = 187.1200 for l-NMA was diminished relative to the same peak in the chromatograms of control samples from which an essential component was omitted (compare orange and red to purple, green and blue). In the complete-reaction samples, a new peak for a species with a change in mass-to-charge ratio (Δm/z) of +16 was also detected (Figure 3C, orange). This new Δm/z +16 peak was barely detectable (< 9 % relative area) in the chromatograms of (i) control samples to which O₂ was not intentionally added (Figure 3C, blue) and (ii) experimental samples from complete reactions with ¹⁸O₂ replacing natural-abundance O₂ (Figure 3C, red). In the latter samples, a peak with Δm/z = +18 instead of +16 developed (Figure 3D, red), confirming that the new peaks are associated with oxygen isotopologs of the oxygenated product. Quantitative analysis of the m/z = 187.1200 chromatograms from the complete-reaction and control samples (as described in the Experimental Procedures) implied consumption of 0.6 (± 0.1) l-NMA per SznF or SznF-Δcupin in the reaction.

Figure 3.

LC-MS analysis of SznF reactions under single turnover conditions demonstrating that l-NMA undergoes oxygenation in association with decay of the peroxo-Fe₂(III/III) complex. Extracted ion chromatograms for the substrate l-NMA (m/z = 187.1200 for [M – H]⁻) are shown for (A) the SznF-wt and (B) the SznF-Δcupin reactions and negative controls. The legend indicates the O₂ isotopolog (natural abundance or ¹⁸O₂) used in complete reactions and the component omitted in the control samples. Extracted ion chromatograms for (C) [¹⁶O]-l-HMA (m/z = 203.1150 for [M – H]⁻) and (D) [¹⁸O]-l-HMA (m/z = 205.1192 for [M – H]⁻) produced in the reactions of the two SznF proteins with the two O₂ isotopologs (red and orange) but not the corresponding control reactions lacking O₂ (blue).

We also directly confirmed reaction of the 629-nm-absorbing complex with both substrates by sequential-mixing SF-Abs experiments. An initial mix of the SznF/Fe(II) solution with O₂ served to accumulate the intermediate to its maximum concentration. The reaction with substrate was then initiated by a subsequent mix with a given concentration of l-NMA (Figure 2C) or l-HMA (Figure 2D). In the former case, the observed rate constants (k₂) associated with the larger amplitude from fits of the decay curves by the equation for two parallel first-order processes exhibited a hyperbolic (i.e., saturating) dependence on [l-NMA] (Figure 2C, inset). This observation is consistent with a mechanism involving fast (rapid-equilibrium) binding of the substrate to the intermediate state followed by rate-limiting N-oxygenation. According to this interpretation, the hyperbolic fit of the plot of k₂ versus [l-NMA] would imply a dissociation constant of 2 (± 0.2) mM for the SznF•l-NMA complex [with the enzyme in the peroxo-Fe₂(III/III) intermediate state] and a rate constant for N-oxygenation of 34 (± 3) s^-1. The kinetics of decay of the absorbing intermediate in the reaction with l-HMA were more complex; a limited supply of this substrate prevented interrogation of a concentration range sufficient to allow this behavior to be confidently modeled. Nevertheless, the acceleration of decay of the 629-nm-absorbing complex by either substrate implies that it is on the reaction pathway for both N-oxygenation reactions. The two possibilities are (i) that this complex itself initiates the N-oxygenations or (ii) that binding of either substrate triggers its conversion to an undetected second intermediate, which then rapidly oxidizes the substrate.

Characterization of the Intermediate by Freeze-Quench Mössbauer Spectroscopy.

We used freeze-quench Mössbauer spectroscopy (FQ-Möss) to gain further insight into the nature of the 629-nm-absorbing intermediate. We initiated the reaction as in the SF-Abs experiments and terminated it by freezing at times ranging from 0.037 s to 5 h (Figures 4, S2).

Figure 4.

Mössbauer spectra, acquired at 4.2 K with a magnetic field of 53 mT applied parallel to the γ-beam, of freeze-quenched samples from the reaction of Fe(II)•SznF with O₂. The vertical bars are the experimental spectra. The heights of the bars reflect the standard deviations of the absorption values during acquisition of the spectra. These bars appear as dots here as a result of the very high signal-to-noise ratio of the data. The colored lines are theoretical spectra illustrating the fractional contributions to the experimental spectra from the Fe(II)•SznF reactant (green), the peroxo-Fe₂(III/III) intermediate (blue), and the Fe₂(III/III) product (red), respectively. Values of the parameters and relative areas of these spectra are provided in Table S2. (A) Spectra of samples freeze-quenched at short reaction times, during formation and decay of the peroxo-Fe₂(III/III) intermediate: (a) anoxic Fe(II)•SznF reactant solution; (b - d) spectra of samples quenched at 37 ms, 450 ms, and 3 s, respectively. (B) Difference spectra: (a) 37 ms – Fe(II)•SznF reactant; (b) 450 ms – 37 ms; and (c) 3 s – 450 ms. These spectra illustrate the changes associated with conversion of the Fe(II)•SznF reactant to the peroxo-Fe₂(III/III) intermediate and Fe₂(III/III) product. (C) Spectra of samples frozen at longer reaction times: (a) 100 s, (b) 10 min, (c) 30 min, (d) 90 min, and (e) 5 h. These spectra demonstrate conversion of the Fe₂(III/III) product cluster to uncoupled high-spin Fe(III) species.

In the low-temperature (4.2 K), weak-field (53 mT oriented parallel to the γ beam) spectra of samples freeze-quenched during the time of formation and decay of the absorbing complex (panel A), the intensity of the quadrupole doublet with δ ~ 1.2 mm/s and ΔE_Q ~ 3.0 mm/s arising from the Fe(II) reactant complex (spectrum a) decreases with increasing reaction time, and two new quadrupole-doublet spectra develop in sequence (spectra b - d). The first new quadrupole doublet exhibits a time dependence similar to that of the 629-nm-absorbing species (blue lines, representing 33, 43, and 24 % of the total absorption area in spectra b, c, and d, respectively; see also Figure S3). Its parameters (δ = 0.60 mm/s and ΔE_Q = 1.00 mm/s), determined from analysis of the [37 ms – Fe(II)•SznF reactant] difference spectrum (panel B, spectrum a), are similar to those associated with synthetic antiferromagnetically coupled, high-spin µ-peroxo-Fe₂(III/III) complexes and analogous complexes formed in FDOs and UndA.^8,14 In particular, the value of δ (0.60 mm/s), which is greater than those generally seen for high-spin Fe(III) species with nitrogen and oxygen coordination,²² is squarely in the range measured for these other peroxide-bridged complexes. The spectroscopic data on the early phase of the SznF reaction thus imply the rapid formation of a µ-peroxo-Fe₂(III/III) intermediate. From comparison of the concentration of this complex as a function of reaction time, determined from analysis of the Mössbauer spectra of the freeze-quenched samples, to the A₆₂₉-versus-time trace from an SF-Abs experiments under precisely the same reaction conditions (Figure S3), a molar absorption coefficient of the complex (ε₆₂₉) of 3.1 (± 0.5) mM⁻¹cm⁻¹ was estimated.

Iron Products from Decay of the Intermediate in the Absence of Substrate.

By a reaction time of 3 s, when the 629-nm feature has partly decayed, the intensity of the quadrupole doublet associated with the µ-peroxo-Fe₂(III/III) complex has partly given way to a new quadrupole-doublet spectrum, and broad magnetically split features extending from ~ −8 mm/s to ~ +8 mm/s have begun to develop. Analysis of the (450 ms – 37 ms) and (3.0 s – 0.45 s) difference spectra yielded the parameters of the quadrupole doublet (δ = 0.52 mm/s and ΔE_Q = 0.44 mm/s; Figure 4B, spectra b and c; see also Figure S2). This lesser value of δ – more usual for high-spin Fe₂(III/III) complexes without bridging peroxides – implies that the third doublet is associated with an antiferromagnetically coupled Fe₂(III/III) complex. The surprisingly small value of ΔE_Q suggests that this product does not have a µ-oxo-Fe₂(III/III) core structure, as is known or thought to form upon decay of the bridged peroxo complexes in reactions of several other non-heme diiron proteins and model complexes.^8,14

At reaction times ≥ 100 s, the changes to the Mössbauer spectrum are characterized by decay of the third quadrupole-doublet component and development of the aforementioned broad, magnetically split features, which are characteristic of high-spin mononuclear Fe(III) species (Figures 4C and S4). By a reaction time of 5 h (spectrum e), the magnetic features contribute most of the absorption area of the Mössbauer spectrum (~ 70 % of the total). The implication is that the Fe₂(III/III) product cluster is not stable in SznF, decaying to uncoupled Fe(III) ions. These Fe(III) products may remain temporarily bound to the protein or rapidly dissociate into solution. Regardless of their location(s) and structure(s), their formation from the oxidized cofactor rationalizes why SznF is isolated following overproduction in E. coli with little iron bound and why initial attempts to solve the structure of the protein with the diiron cofactor in its HDO domain were unsuccessful.

Iron Products from Decay of the Intermediate in the Presence of the l-NMA Substrate.

The spectra of Figure 4 (and Figures S2, S4) report changes in the state of the iron during formation and unproductive decay of the peroxo-Fe₂(III/III) intermediate, in the absence of substrate. We considered that the outcome might be different upon productive decay of the intermediate by substrate N-oxygenation. We addressed this possibility in a sequential-mixing FQ-Möss experiment, in which we allowed the intermediate to accumulate to its maximum concentration and then either freeze-quenched it directly or mixed it with substrate before freezing at different reaction times (Figure 5). As expected on the basis of the sequential-mixing SF-Abs experiment, the quadrupole doublet features of the peroxo-Fe₂(III/III) complex are nearly absent in the spectrum of the sample frozen 0.36 s after mixing with 3.0 mM (final concentration) l-NMA. The features of the difference spectrum (Figure 5C, dotted black line) can be accounted for as the superposition of the quadrupole doublets associated with the peroxo-Fe₂(III/III) intermediate (blue line, −48 % intensity; these features decay) and the Fe₂(III/III) successor complex (red line, 40 % intensity, these feature develop) with the parameters quoted above. In addition, the difference spectrum suggests disappearance of Fe(II) and generation of mono-Fe(III) species. The spectrum of the 30-min sample (Figure 5D, points) is essentially indistinguishable from that of the 30-min sample frozen following decay of the peroxide complex in the absence of substrate (Figure 5D, solid line). These observations suggest that the Fe(III) species formed during and after decay of the peroxo-Fe₂(III/III) intermediate are likely identical, regardless of whether the intermediate decays unproductively or productively.

Figure 5.

Mössbauer spectra, acquired at 4.2 K with a magnetic field of 53 mT applied parallel to the γ-beam, of freeze-quenched samples from the reaction of the μ-peroxo-Fe₂(III/III) complex in SznF with the first substrate, l-NMA. (A) Spectrum of a sample prepared as in Figure 3, panel A, spectrum c (single mix, 450 ms) to verify accumulation of the intermediate. (B) Spectrum of a sample freeze-quenched 360 ms after mixing the solution containing the pre-accumulated intermediate with l-NMA (final concentration of 3.0 mM). (C) Difference spectrum, B - A (vertical bars), plotted with the contributions of the μ-peroxo-Fe₂(III/III) complex that decays (blue line) and the Fe₂(III/III) product that forms (red line) in the reaction with the substrate. The sum of the contributions from the two complexes is also plotted (solid black line). (D) Spectrum of a sample prepared as in B but allowed to react for 30 min before freezing (^.......). The spectrum of the 30-min sample from the reaction in the absence of substrate (Figure 3, panel C, spectrum c) is replotted (——) to facilitate direct comparison.

DISCUSSION

The kinetic and spectroscopic data presented above demonstrate that a µ-peroxo-Fe₂(III/III) complex is an intermediate in both N-hydroxylation steps mediated by SznF on the pathway to SZN. Although intermediates of this general structural type were previously detected in two other N-oxygenases, AurF and CmlI, these enzymes are both FDOs;^{15,16,17,18,19,20} SznF provides the first example of such a complex in an N-oxygenase within the emerging HDO structural superfamily. Recent reports indicate that there are other such HDO N-oxygenases, including on the pathways to azomycin (2-nitroimidazole),²³ gramibactin,^24,25 fragin,²⁶ and pyrazine N-oxide natural products.²⁷ We expect that these enzymes will be found to deploy similar intermediates.

In addition to this protein-structural distinction, characteristics of its substrate(s) and µ-peroxo-Fe₂(III/III) intermediate potentially distinguish the SznF-mediated guanidine hydroxylations from the initial transformations catalyzed by CmlI and AurF. First, the two FDOs both operate on arylamine substrates, which have homolytic N–H-bond dissociation enthalpies (~ 90 kcal/mol) considerably less than those for guanidine derivatives (~ 100 kcal/mol), such as those hydroxylated by SznF.²⁸ Second, the arylamine substrates have pK_a values much less than neutrality and are expected to bind in their unprotonated forms, whereas the l-NMA substrate of the first SznF hydroxylation has a pK_a much greater than 7 (by > 6 units) and would be expected to bind in the guanidinium form.²⁹ Third, the spectroscopic properties (absorption and Mössbauer) of the SznF intermediate are quite different from the nearly matching features of the AurF and CmlI complexes.^{15,16,17,18,19,20} Finally, SznF switches regiochemistry to target two different nitrogen atoms of its substrate via the same intermediate, whereas the peroxide complexes in AurF and CmlI mediate sequential oxidations of a single nitrogen atom.

Might these differences require/reflect different reaction mechanisms? The extensive literature on non-heme diiron enzymes provides two broad mechanistic classes, distinguished by the nature of the initiating intermediate, to be considered for the N-oxygenations mediated by SznF. Early work revealed cases in which μ−1,2-peroxo-Fe₂(III/III) complexes convert to high-valent successors.^30,31 The resultant Fe₂(III/IV) complex, X, in the Ec class Ia ribonucleotide reductase β subunit^32,33 and Fe₂(IV/IV) complex, Q,^34,35,36,37 in soluble methane monooxygenase hydroxylase (sMMOH),^{38,39,40,41,42} both ferritin-like proteins, were shown to remove hydrogen (H•) from the phenolic oxygen of a tyrosine and from methane, respectively. Later studies on FDOs, including toluene/o-xylene monooxygenase hydroxylase (ToMOH), stearoyl acyl carrier protein Δ⁹ desaturase (Δ⁹D), sMMOH with non-native ether substrates,^43,44,45 and, most recently, the arylamine N-oxygenases (AurF and CmlI), invoked direct reaction of peroxide-level complexes with substrates.^15,19 As some known μ−1,2-peroxo-Fe₂(III/III) complexes are quite stable,^46,47 these studies proposed structural changes to this core to enhance reactivity of the peroxide-level intermediate. The μ−1,2-peroxo-Fe₂(III/III) complexes have 600–700-nm absorption features with large molar absorption coefficients of 1.5 – 2.0 mM⁻¹cm⁻¹ and Mössbauer parameters of δ ≥ 0.63 mm/s and |ΔE_Q| > 1 mm/s.^8,14,48 The complexes in ToMOH, AurF and CmlI, were, by contrast, found either to lack a detectable absorption feature (ToMOH) or to have a higher-energy feature (λ_max ~ 500 nm) with diminished absorptivity (≤ 0.5 mM⁻¹cm⁻¹) and to exhibit lesser values of both δ and |ΔE_Q|.^15,18 These differences were, in each case, taken as evidence for protonation of the μ−1,2-peroxide moiety, alteration of peroxide coordination from the canonical μ−1,2 mode to the μ−1,1 or μ-(η¹:η²) bridging mode, or both (Scheme 2A). The protonated/isomerized peroxide-level complexes were proposed to be activated for direct attack^{15,19,37,42,45,49} on the substrates as electrophiles (in ToMOH, AurF, and CmlI), one-electron oxidants (AurF and CmlI),^17,18 or H• abstractors (Δ⁹D)^48,50. The most extensive analysis of intermediate structure and reaction mechanism, for CmlI, converged on a mechanism involving nucleophilic attack of the arylamine on the pendant oxygen atom of a μ−1,1-peroxo-Fe₂(III/III) complex (Scheme 2A, top structure).¹⁹

Scheme 2.

(A) Generalized reaction of a diferrous cofactor with O₂ yielding μ-1,2-peroxo-Fe₂(III/III) precursor and protonated/isomerized successors and (B, C) two possible mechanisms for reaction of the detected (presumptively) μ-1,2-peroxo-Fe₂(III/III) complex in SznF with the guanidinium moiety of l-NMA.

The spectroscopic properties of the SznF intermediate are generally more similar to those of the canonical μ−1,2-peroxo-Fe₂(III/III) complexes than to those of the putatively protonated and/or isomerized adducts in AurF, CmlI, and ToMOH. In particular, the low energy and high absorptivity of its presumptive peroxide-to-Fe(III) charge transfer (LMCT) band suggest that it most likely has the canonical core structure,⁴⁸ although this inference remains to be more firmly established by detailed structural studies (e.g., by resonance Raman and x-ray methods).

In light of the hypothesis (which has been supported by computational studies^17,48,50,51) that protonation should activate the complex and the reasonable presumptions that (i) the dialkylguanidine moiety of l-NMA should initially be protonated and (2) the guanidium should not be appreciably electrophilic or reducing, a set of possible mechanisms would have the μ−1,2-peroxide accepting a proton from the substrate, subsequently (or concertedly) to be attacked by the neutral guanidine (Scheme 2). The nitrogen to be hydroxylated could act as both acid and nucleophile in sequential steps (Scheme 2B). Alternatively, the trigonal symmetry of the guanidinium cation could create a favorable intermediate/transition-state geometry for concerted proton donation by one nitrogen and attack by another (Scheme 2C). In either case, tautomerization and net protonation steps (not depicted) would complete the reaction. A dual role of the substrate as activating Brønsted acid and nitrogenous nucleophile should be testable by kinetic analysis of reactions with appropriately modified l-NMA analogs.

The lability of the oxidized (product) cluster in SznF demonstrated herein, along with previous accounts of cluster instability in UndA and other more recently discovered HDOs,^{23,24,25,26,27,52} raise the possibility that this emerging class of diiron enzymes may be functionally programmed for instability of the oxidized cofactor. This characteristic would distinguish HDOs from the longer-known FDOs, with their stably bound cofactors. Cluster instability could arise from the flexibility of ligand-contributing core helices and solvent-accessibility of the metal-binding core suggested by the structures of HDOs that have been solved to date.^11,12 Such flexible coordination could enable more rapid reconfiguration of substrate-binding sites, potentially to accommodate direct coordination of the substrate to the cofactor, as seen in UndA.^9,11 It could also facilitate product release, particularly in cases in which the N-hydroxylated products avidly bind Fe(II) or Fe(III). In these scenarios, the HDOs may have emerged recently, imparting novel reactivities to support diverse secondary metabolic pathways.^{23,24,25,26,27,52} Whether the now repeatedly observed cofactor lability is an adaptive, functional feature of the emerging HDO superfamily or an in vitro artifact reflecting the absence of stabilizing component proteins remains to be established.

EXPERIMENTAL PROCEDURES

Materials.

Isopropyl-β-D-thiogalactopyranoside (IPTG) was purchased from DOT Scientific. 4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), glycerol, and imidazole were purchased from Fisher Scientific. Kanamycin was purchased from Teknova. Sodium chloride (NaCl), sodium hydroxide (NaOH), ammonium chloride (NH₄Cl) and hydrochloric acid (HCl) were purchased from EMD Millipore. Ni(II)-nitrilotriacetic acid agarose (Ni-NTA) resin was purchased from Qiagen. 1H-pyrazole-1-(N-methylcarboxamidine)hydrochloride was purchased from Alfa Aesar. All other chemicals used were purchased from Sigma Aldrich. Escherichia coli (Ec) DH5α and BL21(DE3) competent cells were purchased from Novagen. ⁵⁷Fe⁰ was purchased from ISOFLEX, USA.

Preparation of l-HMA.

The procedure for preparing l-HMA was modified slightly from that reported in the published work.¹ The nitrone (Scheme S1, left) (500 mg, 1.25 mmol) was first dissolved in 20 mL dimethylformamide (DMF), and hydroxylamine hydrochloride (NH₂OH•HCl, 95.9 mg, 1.38 mmol, 1.1 equiv) was added in one portion. After 1 h of stirring at room temperature under N₂, 1H-pyrazole-1-(N-methylcarboxamidine) hydrochloride (221 mg, 1.38 mmol, 1.1 equiv) was added in one portion, and triethylamine (NEt₃, 348.0 μL, 2.5 mmol, 2.00 equiv) was added dropwise. The resulting mixture was stirred overnight under N₂ at room temperature. After 16 h of stirring at room temperature, the solution was concentrated to ~ 5 mL under a stream of N₂. The mixture was applied directly to a C18 Sep-Pak cartridge (Waters) to remove polar impurities. After the cartridge was washed with 50 mL water + 0.1% TFA (trifluoro-acetic acid), Cbz-l-HMA was eluted by 50 mL acetonitrile + 0.1% TFA. The organic solution containing Cbz-l-HMA was concentrated in vacuo and lyophilized overnight to remove water.

The Cbz group was removed by dissolving crude Cbz-l-HMA in 10 mL of 10% (v/v) thioanisole in TFA. After the mixture was stirred overnight at room temperature, TFA was removed in vacuo. A 10 mL aliquot of toluene was added to the resulting crude oil, and the solution was concentrated in vacuo again. The crude oil was first triturated with diethyl ether then dissolved in 5 mL of water. The aqueous solution containing l-HMA was then washed with 2 x 5 mL of diethyl ether and 2 x 5 mL of dichloromethane. The aqueous solution containing l-HMA was then applied to a C18 Sep-Pak cartridge (Waters) to remove non-polar impurities. l-HMA was eluted with 20 mL water + 0.1% TFA. This aqueous solution was lyophilized to dryness to afford l-HMA as a red oil. The oil was triturated with ice-cold acetonitrile, and the precipitate was filtered to afford 43 mg of l-HMA (7.9% yield from nitrone). High-resolution mass spectrum and ¹H NMR data matched previous reports.¹

Overexpression and Purification of SznF in M9 Medium with Mn(II) Supplementation.

A pET29a vector encoding N-terminally His₆-tagged Sa SznF¹ was used to transform Ec BL21(DE3) cells. Successful transformants were selected on LB-agar plates supplemented with kanamycin (50 μg/mL) and incubated at 37 °C. A single colony was used to inoculate a 5 mL LB starter culture containing 50 μg/mL kanamycin. The starter culture was grown at 37 °C with shaking at 180 rpm for 6 h. The cells were harvested by centrifugation at 10,000 x g for 5 min and resuspended in 5 mL of M9 medium.⁵³ A 1 mL aliquot of this suspension was used to inoculate a 250 mL M9 starter culture supplemented with 50 μg/mL kanamycin and 50 μM MnCl₂. The starter culture was incubated at 37 °C overnight (16 h) with shaking at 180 rpm. Aliquots of 25 mL of a starter culture were used to inoculate 1 L cultures of M9 medium supplemented with 50 μg/mL kanamycin and 50 μM MnCl₂. The cultures were grown at 37 °C with shaking at 180 rpm to an OD₆₀₀ of ~ 0.6 ‒ 0.8. After cooling on ice for 30 min with an additional supplement of 200 μM MnCl₂, SznF overexpression was induced with 0.5 mM IPTG (final concentration) and grown at 18 °C overnight (16–18 h) with shaking at 180 rpm. Cells were harvested by centrifugation at 8000g and flash frozen in liquid N₂. A typical growth yielded 25 g of frozen cell paste.

The frozen cells were resuspended in buffer A [100 mM sodium HEPES, 10 mM imidazole, 5% glycerol (v/v), pH 7.5, supplemented with 80 μg/ml PMSF] at ratio of 5 mL per g of cells. The suspended cells were disrupted via sonication on ice (QSonica 750 W, 20 kHz, 10 s pulse, 30 s off, 60% amplitude, 7 min 30 s total sonication time). The resulting lysate was clarified by centrifugation at 22,000 x g. This step and all subsequent protein purification steps were performed at 4 °C. The supernatant was applied to Ni(II)-nitrilotriacetate (NTA) agarose resin (1 mL resin per 5 mL lysate) and washed with 6 column volumes (CV) of buffer A. Bound proteins were eluted with buffer B [100 mM HEPES, 250 mM imidazole, 5% glycerol (v/v), pH 7.5]. Fractions containing SznF, as determined by SDS-PAGE gel analysis, were pooled and concentrated in a 30 KDa molecular weight cutoff filtration device (PALL Corporation). To remove adventitiously bound metal ions, the protein was dialyzed overnight into 20 mM sodium HEPES, 50 mM NaCl, 5% (v/v) glycerol, 10 mM EDTA, pH 7.5. To remove EDTA, the protein was dialyzed against 20 mM sodium HEPES, 50 mM NaCl, 5% (v/v) glycerol, pH 7.5 for at least 4 h, followed by desalting with a pre-packed PD-10 column (GE Healthcare). Metal content (Table S1) was determined at the Penn State Institute of Energy and the Environment Laboratory on a Perkin-Elmer Optima 5300 inductively coupled plasma emission spectrometer. Samples were prepared both prior to and after EDTA treatment by dissolution of protein in 3.5% nitric acid. For other experiments, protein samples were rendered anoxic on a Schlenk line after purification. Three sets of ten evacuation and purge cycles with argon gas were performed, and the resulting anoxic samples were stored under liquid N₂. Protein concentrations were determined by UV-visible absorbance at 280 nm using a calculated molar absorption coefficient (ε_280nm) of 70,500 M⁻¹ cm⁻¹ and a predicted molecular weight of 55.7 kDa.⁵⁴

Single turnover LC-MS activity assay.

The reaction buffer consisted of 20 mM sodium HEPES, 50 mM NaCl, 5% (v/v) glycerol, pH 7.5. An anoxically prepared 25 μL sample containing SznF (0.3 mM), l-NMA (0.6 mM), and ferrous ammonium sulfate (0.9 mM) was mixed with an additional 25 μL of O₂-saturated buffer (~1.8 mM O₂). In samples containing ¹⁸O₂ or from which O₂ was omitted, 25 μL of ¹⁸O₂-saturated buffer or anoxic buffer (respectively) was substituted. Two additional controls were carried out: in one, ferrous ammonium sulfate was omitted; in the second, SznF(-Δcupin) was omitted. After 30 min at 22 °C, the reactions were terminated by addition of 50 μL of LC-MS grade methanol. Each sample was incubated on ice for 30 min, centrifuged for 10 min at 21,000 x g at 4 °C to remove precipitated protein, and analyzed by normal phase LC-MS.

Samples were further diluted 1:1 with LC-MS grade methanol before analysis, which was performed on an Agilent 1260 Infinity series LC system coupled to an Agilent 6530 Accurate Mass Quadrupole Time-of-Flight (qTOF) mass spectrometer with an electrospray ionization (ESI) source. A Cogent 4 Diamond Hydride column (4 μ, 100 Å, 150 mm x 3.0 mm) was used for chromatographic separation. Samples were analyzed during a gradient from 10% solution B (0.1% formic acid in water) and 90% solution A (0.1% formic acid in acetonitrile) to 70% solution B over 19 minutes at a flow rate of 0.5 mL/min. Mass spectra were recorded under negative ionization mode, with a mass range of 100 to 1700 m/z, 1 spectra/s, 3500 V capillary voltage, 35 psi nebulizer pressure, 8 L/min drying gas (N₂), and gas temperature at 300 °C.

Exact ion chromatograms (EICs) were extracted with a mass window of 10 ppm for l-NMA, l-HMA, and [¹⁸O]-l-HMA. Both the [M – H]⁻ ion and [M – C₄H₂N₂]⁻ ion (representing the loss of a guanidine-derived species by fragmentation between N^δ and the central carbon) were monitored. In quantification of the l-NMA substrate consumed in the SznF(-Δcupin) reactions, the areas under the EICs for both ions were summed to improve precision. As expected, the six control samples [–O₂, –Fe(II), and –enzyme samples paired with the SznF-WT and SznF-Δcupin complete-reaction samples], in which no substrate should have been consumed and no product generated, gave peaks for the l-NMA substrate with similar areas (5.1% standard error) and no peak for the product (or, for the case of the –O₂ controls, a very small peak, as a result of incomplete removal of atmospheric O₂). The areas of the substrate-associated peaks for the six control samples were therefore averaged to provide the reference area for the complete-reaction samples. Also as expected – in light of the prior demonstration that the Δcupin variant is fully active for the N-hydroxylation steps¹ – the four complete-reaction samples (SznF-WT and SznF-Δcupin with ¹⁶O₂ and ¹⁸O₂) gave similar areas for the substrate-associated peaks (2.4% standard error) that were all significantly less (by a mean of 30%) than the mean area of the substrate-associated peaks for the six control samples. The ratio of the mean areas for the complete-reaction and control samples thus yielded an estimate of the fraction of l-NMA remaining after the reaction (0.7), and the quantity consumed was calculated from this fraction and the known initial equivalency of l-NMA substrate [(1-0.7) x 2 equiv = 0.6 (± 0.1) equiv consumed].

Stopped-Flow Absorption (SF-Abs) Spectroscopy.

All SF-Abs experiments were carried out at 5 °C in an Applied Photophysics Ltd. (Leatherhead, UK) SX20 stopped-flow spectrophotometer housed in an MBraun anoxic chamber and equipped with a photodiode-array detector. The stopped-flow apparatus was configured for either single or sequential mixing and a path length of 1 cm (except in one case, which is noted). Time-dependent absorption spectra (1000 per trial) were acquired with a logarithmic time base.

In the single-mixing SF-Abs experiments, an anoxic solution containing 0.30 mM apo SznF (wild-type or Δcupin variant) and 0.90 mM (3 molar equiv) Fe(II) in 20 mM sodium HEPES buffer (pH 7.5) containing 50 mM sodium chloride and 5% (w/v) glycerol was mixed with an equal volume of the same buffer saturated with O₂ (~ 1.8 mM). In some experiments, either l-NMA or l-HMA was also included with the protein at an initial concentration of 1.0 mM, giving 0.50 mM after mixing. In the experiments with l-HMA, an increased sodium HEPES concentration (50 mM) was used, with the other buffer components and pH remaining unchanged. The A_629nm-versus-time trace from each of these experiments was characterized by a biphasic rise and monophasic decay, with nearly equal initial and final absorbance values. In the regression analysis of these traces (solid lines in the figures), we assumed them to reflect accumulation and decay of a single intensely absorbing intermediate state (I in Scheme 3) in a two-step reaction sequence with approximately isosbestic reactant (R) and product (P) states.

Scheme 3.

Kinetic model used in regression analysis of the SF-Abs traces.

Both steps of this sequence were assumed to be irreversible, and, because O₂ was initially present in 6-fold excess over the protein, formation of the intermediate was treated as a pseudo-first-order process with effective rate constant k₁. The biphasic rise phase was interpreted to imply that the Fe(II)•SznF reactant complex actually consists of two forms (R and R’), which combine with O₂ with different effective pseudo-first-order rate constants (k₁ and k₁’) to form I. The equation derived from this model (Eq 1) was used to fit the kinetic traces in Figure 2B.

$$A_t=A_0+\Delta A\left(\frac{k_1}{k_2-k_1}\right)\left(e^{-k_{1}t}-e^{-k_{2}t}\right)+\Delta A^{\prime }\left(\frac{k_1^{\prime }}{k_2-k_1^{\prime }}\right)\left(e^{-k_1^{\prime }t}-e^{k_{2}t}\right)$$ (1)

In the sequential-mixing SF-Abs experiments, an anoxic solution containing 0.60 mM apo SznF and 1.8 mM Fe(II) in the same buffer described above was mixed with an equal volume of the O₂-saturated buffer, and this reaction solution was incubated for 0.56 s to allow the intermediate to accumulate maximally before it was mixed with an equal volume of an anoxic solution of l-NMA or l-HMA, giving the final substrate concentrations indicated in Figure 2, panels C and D. For the case of l-HMA, the higher sodium HEPES concentration (50 mM) was again used. In these sequential-mixing experiments, formation of the absorbing intermediate occurred in the unmonitored initial incubation, and the traces obtained reflect only decay of the intermediate either in the absence or in the presence of substrate. The trace from the reaction without substrate is nearly monophasic and was fit as a single-exponential decay (Eq. 2). In the presence of either substrate, two distinct phases of decay invariably became evident in attempts to fit the data by Eq 2. These traces were fit by Eq. 3, which corresponds to decay by two parallel first-order processes with rate constants k₂ and k₂’. For the case of l-NMA, the larger of the two rate constants, which was invariably associated with the larger of the two amplitudes, was plotted versus l-NMA concentration in the inset to Figure 2C. These data were then fit by the equation for a hyperbola (solid line), yielding the parameters (given in the Results section) that can be equated (with certain assumptions) to the dissociation constant (K_d) for the peroxo-Fe₂(III/III)-SznF•l-NMA complex and the rate constant for decay of the intermediate by hydroxylation of N^δ.

$$A_t=A_0+\Delta A_1•e^{-k_{2}t}$$ (2)

$$A_t=A_0+\Delta A_1•e^{-k_{2}t}+\Delta A_2•e^{-k_2^{\prime }t}$$ (3)

Preparation of Freeze-Quench Mössbauer (FQ-Möss) Samples.

General methods for preparation of freeze-quenched Mössbauer samples have been published previously.⁵⁵ To begin these experiments, an anoxic solution of 2 mM SznF and 6 mM ⁵⁷Fe(II) was first prepared. An O₂-stable, acidic, concentrated stock solution of ⁵⁷Fe(II), prepared from commercial ⁵⁷Fe⁰ as previously described,⁵⁶ was diluted in the anoxic chamber from ~ 500 mM to 50 mM by mixing with 1 M sodium HEPES, pH 7.5. The proper volume of this dilute stock solution was added to an anoxic solution of apo SznF (wild-type or Δcupin) in 20 mM sodium HEPES buffer (pH 7.5) containing 50 mM NaCl and 5% (v/v) glycerol. This Fe(II)•SznF reactant solution was mixed at 5 °C with an equal volume of the same buffer that had been saturated with O₂ (~ 1.8 mM). In the single-mixing experiments, the resultant solution was allowed to incubate for the varying reaction times indicated in the figures and then frozen by injection into cold (−150 °C) 2-methylbutane (for reaction times from milliseconds to tens of seconds) or by pipetting into a Mössbauer cell cooled on a metal block that was in contact with liquid N₂ (for reaction times of minutes to hours). In the sequential-mixing experiments, the reaction solution was incubated for 0.44 s to accumulate the intermediate before being mixed with 0.25 equivalent volume of 15 mM l-NMA in the same buffer, giving 3 mM l-NMA in the final reaction. This solution was further incubated for either 350 ms or 30 min before being freeze-quenched or manually frozen, respectively.

Mössbauer spectroscopy

Mössbauer spectra were recorded on a spectrometer from SEECO (Edina, MN) equipped with a Janis SVT-400 variable-temperature cryostat. The reported isomer shift is relative to the centroid of the spectrum of α-iron metal at room temperature. External magnetic fields were applied parallel to direction of propagation of the γ radiation. Simulations of the Mössbauer spectra were carried out using the WMOSS spectral analysis software from SEECO (www.wmoss.org, SEE Co., Edina, MN).