Spectroscopic Evidence for Electronic Control of Heme Hydroxylation by IsdG

Abstract

Staphylococcus aureus IsdG catalyzes a unique trioxygenation of heme to staphylobilin, and the data presented in this article elucidate the mechanism of the novel chemical transformation. More specifically, the roles of the second-sphere Asn and Trp residues in the monooxygenation of ferric−peroxoheme have been clarified via spectroscopic characterization of the ferric−azidoheme analogue. Analysis of UV/vis absorption data quantified the strength of the hydrogen bond that exists between the Asn7 side chain and the azide moiety of ferric−azidoheme. X-band electron paramagnetic resonance data were acquired and analyzed, which revealed that this hydrogen bond weakens the π-donor strength of the azide, resulting in perturbations of the Fe 3d based orbitals. Finally, nuclear magnetic resonance characterization of ¹³C-enriched samples demonstrated that the Asn7···N₃ hydrogen bond triggers partial porphyrin to iron electron transfer, resulting in spin density delocalization onto the heme meso carbons. These spectroscopic experiments were complemented by combined quantum mechanics/molecular mechanics computational modeling, which strongly suggested that the electronic structure changes observed for the N7A variant arose from loss of the Asn7···N₃ hydrogen bond as opposed to a decrease in porphyrin ruffling. From these data a fascinating picture emerges where an Asn7···N₃ hydrogen bond is communicated through four bonds, resulting in meso carbons with partial cationic radical character that are poised for hydroxylation. This chemistry is not observed in other heme proteins because Asn7 and Trp67 must work in concert to trigger the requisite electronic structure change.

Graphical Abstract

INTRODUCTION

Noncanonical heme oxygenases are a fundamentally new chapter in the rich inorganic chemistry of heme iron and molecular oxygen. From an inorganic standpoint, hemoglobin and myoglobin are classic examples of heme−oxygen chemistry where molecular oxygen binds reversibly to protein-bound ferrous heme to facilitate oxygen transport and storage. Monooxygenases activate heme-bound dioxygen and insert a single oxygen atom into an organic substrate, as exemplified by the hydroxylation of an unactivated alkane by cytochrome P₄₅₀.¹ Dioxygenases are yet another class of heme-dependent enzymes that incorporate both atoms of heme-bound dioxygen into an organic substrate. Another interesting area of heme−oxygen chemistry is heme oxygenases; these enzymes catalyze the self-oxygenation of heme by heme-bound dioxygen. Canonical heme oxygenases, such as those found in eukaryotes, convert heme to a dioxygenated tetrapyrrole, biliverdin.² On the other hand, noncanonical heme oxygenases convert heme to trioxygenated tetrapyrroles such as staphylobilin and mycobilin (Figure 1).³ Since these trioxygenated products are unique, their formation must involve novel heme−oxygen chemistry.

Figure 1.

IsdG degrades heme to the trioxygenated tetrapyrrole product, staphylobilin, via ferric−peroxoheme and meso-hydroxyheme intermediates (top). The goal of this work is to elucidate the chemical mechanism for the ferric−peroxoheme to meso-hydroxyheme conversion. In order to achieve this goal, selectively ¹³C enirched heme was biosynthesized from [5-¹³C]-δ-aminolevulinic acid (bottom).

Previous research by several groups has identified reaction intermediates and catalytically essential amino acids for the noncanonical heme oxygenase IsdG, which converts heme to staphylobilin.^4,5 Biochemical assays and density functional theory modeling of Staphylococcus aureus IsdI, a member of the IsdG enzyme family, revealed that IsdG proceeds through a ferric−peroxoheme intermediate that is similar to the “compound 0” intermediate of monooxygenases and canonical heme oxygenases.⁶ Later, time-resolved mass spectrometry and UV/vis absorption (UV/vis Abs) spectroscopy revealed that this ferric−peroxoheme species is converted to a monooxy-genated meso-hydroxyheme intermediate (Figure 1).⁷ At least three amino acids are required for IsdG-catalyzed conversion of heme to staphylobilin via these intermediates: Asn7, Trp67, and His77.⁸ His77 is a first-sphere residue that forms a coordinate covalent Fe−N bond with the heme iron (PDB ID 2ZDO),⁹ whereas Asn7 and Trp67 are second-sphere residues that interact with the His−heme−O₂ moiety via nonbonding interactions. Previous studies have revealed that steric interactions between the second-sphere Trp and the porphyrin ring of heme trigger an out-of-plane ruffling deformation that has been correlated to a heme electronic structure change and an increased heme degradation rate in S. aureus IsdI.^10,11 On the other hand, significantly less is known regarding the function of the catalytically essential Asn7 residue.

Thus, the purpose of this study was to elucidate the function of the second-sphere Asn residue in noncanonical heme oxygenases. Asn7 is absolutely essential for enzymatic turnover since no partial or complete heme degradation has been reported for any variant of IsdG where the conserved Asn7 residue has been replaced by another amino acid.⁸ A previously reported spectroscopic characterization of S. aureus IsdG deduced that a hydrogen bond exists between Asn7 and the iron-ligating α-oxygen of a distal peroxo ligand.¹² This hydrogen bond orients the bent peroxo ligand along the β/δ meso carbon axis of the porphyrin ligand and perturbs the heme electronic structure. An intriguing hypothesis that arose from density functional theory modeling of this hydrogen-bonding interaction was that there may be long-range electronic communication between Asn7 and the porphyrin ligand that induces partial electron transfer from porphyrin to iron, resulting in spin density delocalization onto the heme meso carbons without an electronic ground state change from ²E_g to ²B_2g. These heme carbons are the hydroxylation site going from ferric−peroxoheme to meso-hydroxyheme (Figure 1), and the introduction of partial cationic radical character would be expected to promote attack by the anionic oxygen radical produced by homolytic cleavage of the ferric−peroxoheme O−O bond. Nevertheless, this computational prediction has not yet been experimentally tested because the influence of Asn7 on the spin density of the heme meso carbons has not been measured. Unfortunately, the ferric−peroxoheme form of IsdG is an unstable reactive intermediate;⁶ therefore, an analogue of this species must be used for indepth spectroscopic interrogation. The ferric−azidoheme form of IsdG has previously been shown to be an ideal analogue since,¹² similarly to peroxide, azide is both a π-donor ligand and a hydrogen bond accepting ligand via its iron-ligating α-nitrogen. This motivated us to assess the effect of Asn7 on the porphyrin spin density in the ferric−azidoheme analogue of ferric−peroxoheme.

Here, a multifaceted spectroscopic and computational approach was employed to elucidate the functional role of the second-sphere Asn residue in IsdG enzymes. UV/vis Abs spectroscopy was used to measure the dissociation constant (K_d) values for azide dissociation from wild-type (WT) and N7A forms of azide-inhibited IsdG (IsdG−heme−N₃) and gain insight into the strength of the Asn···N₃ hydrogen bond. Next, X-band electron paramagnetic resonance (EPR) spectroscopy was utilized to determine how this hydrogen bond perturbs the iron electronic structure. Finally, ¹H and ¹³C nuclear magnetic resonance (NMR) spectroscopy was employed to assess the effect of the Asn···N₃ hydrogen bond on the porphyrin electronic structure. To aid in the assignment of the particularly relevant heme meso ¹³C resonances, selectively ¹³C enriched heme was prepared biosynthetically (Figure 1).¹³⁻¹⁵ These spectroscopic characterizations were complemented by experimentally validated hybrid quantum mechanics/molecular mechanics (QM/MM) calculations to ascertain whether the electronic structure changes arose from the Asn···N₃ hydrogen bond and/or Asn7-induced changes to porphyrin ruffling. The implications of these data with respect to the ferric−peroxoheme electronic structure and the fundamental requirements for the novel oxygenation reaction catalyzed by IsdG are discussed.

EXPERIMENTAL SECTION

Unless otherwise noted, all materials in this work were purchased from Fisher Scientific and used without further purification.

Expression and Purification of IsdG.

The cloning, expression, and purification of long-linker WT and N7A IsdG from pET-15b (Amp^r, Novagen),^12,16,17 and S219V tobacco etch virus (TEV) protease from PRK793 (Amp^r),¹⁸ have been previously described. Short linker constructs of WT and N7A IsdG lacking a 12 amino acid N-terminal linker between the TEV protease cleavage site and the start of the isdG gene were prepared using the QuikChange Lightning site-directed mutagenesis kit (Agilent). The DNA primers used for the deletion mutations were purchased from the Midland Certified Reagent Company (Table S1). DNA sequencing performed by the University of Vermont Cancer Center DNA analysis facility confirmed deletion of the 12 amino acid linker (Tables S2 and S3). Prior to protein expression and purification, pET-15b vectors encoding short-linker constructs of WT and N7A IsdG were transformed into BL21-GOLD (DE3) cells (Stratagene).

Short-linker WT IsdG was expressed as previously described for long-linker WT IsdG.^16,17 Several modifications to this procedure were necessary in order to obtain satisfactory yields of short-linker N7A IsdG. Following overnight growth in Luria−Bertani medium with 100 μg/mL of ampicillin at 37 °C, 10 mL were transferred into 1 L of Terrific Broth containing 100 μg/mL of ampicillin and supplemented with a cocktail of trace metals.¹⁹ These cells were grown at 37 °C and 225 rpm using a Thermo Scientific MaxQ 5000 floor-model shaker for 24 h without induction and harvested via centrifugation at 4 °C and 8000g for 10 min using a Thermo Scientific Sorvall Legend XT centrifuge equipped with a FiberLite F14–6×250 LE rotor.

Short-linker WT IsdG was purified as previously described for long-linker IsdG,^12,17 with some minor modifications. Filtered lysate was loaded onto a 5 mL, nickel-charged HiTrap Chelating HP column (GE Healthcare) equilibrated with 50 mM Tris pH 7.4 and 150 mM NaCl using an Äkta Pure fast protein liquid chromatography (FPLC) system (GE Healthcare). The column was washed with a linear gradient of imidazole from 0 to 80 mM prior to elution of WT IsdG with 250 mM imidazole. The N-terminal His₆ tag was removed as previously described for IsdI.^17,20,21 This procedure yielded untagged, short-linker WT IsdG in >99% purity as determined by SDS-PAGE gel electrophoresis (Figure S1).

Several modifications to this procedure were necessary to obtain pure, short-linker N7A IsdG from the uninduced growth described above. The cell pellet from 1 L of growth was resuspended in 80 mL of 50 mM Tris pH 7.4 and 150 mM NaCl with 5 mM phenylmethanesulfonyl fluoride (PMSF, Pierce) and 25 units of DNase I (Pierce). Cells were lysed via sonication in the presence of 0.1 mg/mL of lysozyme using a Branson Sonifier S-450A instrument at 60% duty cycle, 6 output control for 1 min on, 1 min off over a total of 12 min. The lysate was centrifuged at 4 °C and 15000g for 30 min, and then the supernatant was transferred to a clean tube and centrifuged for an additional 1 h. The supernatant was filtered using a 0.45 μm membrane (Millipore), loaded onto a HiTrap Chelating HP column, and purified as described above for short-linker WT IsdG, yielding partially heme bound, short-linker N7A IsdG. Heme was removed from N7A IsdG by incubating the partially heme-bound protein with 1.1 equiv of apomyoglobin, prepared as described previously,^22,23 at 4 °C for 3 h while dialyzing the mixture against 50 mM Tris pH 7.4 and 150 mM NaCl. The dialyzed mixture was loaded onto a 5 mL, nickel-charged HiTrap Chelating HP column equilibrated with 50 mM Tris pH 7.4 and 150 mM NaCl using an Äkta Pure FPLC system (GE Healthcare). The column was washed with a linear gradient of imidazole from 0 to 80 mM prior to elution of short-linker N7A IsdG with 250 mM imidazole. The apomyoglobin extraction reduced the fraction of heme-bound N7A IsdG 7-fold, as estimated by the A_Soret/A₂₈₀ intensity ratio. The N-terminal His₆ tag was removed as described above for WT enzyme, resulting in >99% pure untagged, short-linker N7A IsdG as determined by SDS-PAGE gel electrophoresis (Figure S2).

Expression and Purification of Cytochrome b₅.

A pET11a (Amp^r) plasmid encoding rat liver outer mitochondrial membrane cytochrome b₅ (OM cyt b₅) was a gift from Mario Rivera (University of Kansas). OM cyt b₅ was expressed as described previously,¹³⁻¹⁵ except a different minimal medium recipe, supplemented with [5-¹³C]-δ-aminolevulinic acid (Cambridge Isotope Laboratories), was used to improve expression.¹⁹

OM cyt b₅ was purified as described previously,¹³⁻¹⁵ with minor modifications. Filtered lysate was loaded onto a HiTrap Q HP 5 mL column equilibrated with 100 mM Tris pH 8.0 using an Äkta Pure FPLC system (GE Healthcare). The column was washed with a linear gradient from 0 to 200 mM NaCl, and OM cyt b₅ eluted between 75 and 100 mM NaCl. Purified protein was concentrated using Amicon stirred-cells (Millipore) prior to exchange into 100 mM sodium phosphate (NaP_i) pH 7.0 using a PD-10 desalting column (GE Healthcare). Finally, selectively ¹³C labeled heme (¹³C-heme) was extracted from OM cyt b₅ as previously described.^24,25

Circular Dichroism (CD) Spectroscopy.

CD spectroscopy was used to compare the secondary structures of long-linker and short-linker WT IsdG. Both constructs of WT IsdG were exchanged into 10 mM potassium phosphate (KP_i) pH 7.4 using PD-10 desalting columns and loaded into 2 mm path length quartz cuvettes (Starna). CD spectra were acquired from 260 to 190 nm with a scan speed of 20 nm/min, a bandwidth of 1.0 nm, a digital integration time of 8 s, and a data pitch of 0.5 nm using a Jasco J-815 spectropolarimeter.

UV/Vis Abs Spectroscopy.

Molar extinction coefficients for several forms of short-linker IsdG were determined using established assays. The extinction coefficients for short-linker, WT and N7A heme-bound IsdG (IsdG−heme) were determined as previously described for long-linker IsdG−heme,¹⁷ using the pyridine hemochrome method.²⁶ These measurements resulted in extinction coefficients of ε₄₁₁ = 92.9 mM⁻¹ cm⁻¹ and ε₃₉₉ = 94.1 mM⁻¹ cm⁻¹ for short-linker WT and N7A IsdG−heme, respectively. The extinction coefficients for azide-inhibited IsdG (IsdG−heme−N₃) were determined using a procedure similar to that previously described for long-linker IsdG.¹² Briefly, enough NaN₃ was added to an IsdG−heme solution of known concentration to reach a final azide concentration of 100 mM, and UV/vis Abs spectra were acquired from 700 to 300 nm using a Cary 100 Bio spectrophotometer with a scan rate of 600 nm/min, a 1.0 nm data interval, and a 0.1 s integration time. These experiments revealed extinction coefficients of ε₄₁₈ = 86.0 mM⁻¹ cm⁻¹ and ε₄₁₇ = 92.1 mM⁻¹ cm⁻¹ for short-linker WT and N7A IsdG−heme−N₃, respectively. The extinction coefficients measured here for short-linker IsdG are within 5% of those previously measured for the long-linker enzyme.⁷

UV/vis Abs detected titrations of azide into IsdG−heme were analyzed in order to determine the K_d value for the azide ligand. WT and N7A IsdG−heme in 50 mM Tris pH 7.4 and 150 mM NaCl were prepared as described above and combined with an equal volume of NaN₃ in the same buffer to bring the final concentration of IsdG−heme to 3.75 μM with the azide concentration ranging between 0 and 200 mM. The UV/vis Abs intensity at 425 nM for a mixture of IsdG−heme and IsdG−heme−N₃ depends upon eq 1

$$\begin{gathered} A_{425}=\frac{\left({\epsilon }_{\text{IsdG}-\text{heme}-{\text{N}}_3}-{\epsilon }_{\text{IsdG}-\text{heme}}\right)\left[\text{IsdG}-{\text{heme}}_{\text{T}}\right]\left[{\text{N}}_{3,\text{T}}\right]}{K_{\text{d}}+\left[{\text{N}}_{3,\text{T}}\right]} \\ +{\epsilon }_{\text{IsdG}-\text{heme}}\left[\text{IsdG}-{\text{heme}}_{\text{T}}\right] \end{gathered}$$ 1

where ε_{IsdG−heme−N3} and ε_IsdG−heme are the molar extinction coefficients for IsdG−heme−N₃ and IsdG−heme, respectively, at 425 nm, [IsdG−heme_T] is the total IsdG concentration, and [N_3,T] is the total azide concentration. [IsdG−heme_T] was constrained to its known value, and A₄₂₅ as a function of [N_3,T] was fit to eq 1 using GraphPad Prism 7.0a in order to determine the K_d value and its standard error. The molar extinction coefficients derived from this analysis were compared to those determined using the pyridine hemochrome method,²⁶ and both data sets agreed within 2.5 mM⁻¹ cm⁻¹. A complete derivation of eq 1 is available in the Supporting Information.

EPR Spectroscopy.

Long-linker WT and N7A IsdG−heme were prepared as previously described.¹² Purified WT and N7A IsdG−heme were exchanged into 125 mM KP_i pH 7.4 using PD-10 desalting columns, and enough sodium azide was added to reach a final concentration of 100 mM. For each IsdG−heme−N₃ sample, the protein concentration was increased to 250 μM using Amicon ultra centrifugal filters and the sample was transferred to a quartz EPR tube (Wilmad) and flash-frozen in liquid nitrogen.

X-band (9.384 GHz) EPR spectra were acquired on a Bruker EleXsys E-500 spectrometer with the sample temperature maintained at 10 K by an Oxford continuous flow liquid helium cryostat. EPR spectra were collected with 6.3−63 μW of microwave power, a 100 kHz modulation frequency, 10 G modulation amplitude, and a time constant of 5.12 ms. EasySpin 5.2.21 was used to simulate the EPR spectra and extract spin Hamiltonian parameters at three microwave powers (Table S4).²⁷ Complete parameter sets for simulations are available in the Supporting Information (Tables S5 and S6). A ligand field theory based analysis method was used to determine the axial (Δ/λ) and rhombic (V/λ) terms for the low-spin species from experimentally derived g values.²⁸

NMR Spectroscopy.

Long-linker IsdG−heme was prepared as described above, and short-linker IsdG−heme with ¹³C-heme (IsdG−¹³C-heme) was prepared by adding 1.0 mL of ¹³C-heme in dimethyl sulfoxide to 75 mL of short-linker IsdG in 50 mM Tris pH 7.4 and 150 mM NaCl, yielding a final molar ratio of 0.8:1 (¹³C-heme:IsdG). IsdG−heme and IsdG−¹³C-heme were then exchanged into 22 mM NaP_i pH 7.4 using PD-10 desalting columns (GE Healthcare). Enough NaN₃ was added to reach final concentrations of 110 mM, the volumes were reduced to 540 μL using Amicon stirred cells (Millipore), and 60 μL of D₂O (Cambridge Isotope Laboratories) was added to yield IsdG−heme and IsdG−¹³C-heme samples in 20 mM NaP_i pH 7.4, 100 mM azide, and 10% D₂O (v/v). The final concentrations of IsdG−heme−N₃ and azide-inhibited IsdG−¹³C-heme (IsdG−¹³C-heme−N₃) ranged from 1.0 to 1.5 mM.

¹H NMR spectra were acquired at 25 °C using a Varian Unity Inova 500 MHz instrument equipped with an inverse triple-resonance probe. ¹H NMR experiments utilized a 3 s recycle time, a 1 s acquisition time with presaturation of the water ¹H resonance during the relaxation delay, and a 105 kHz sweep width. All ¹H NMR spectra were indirectly referenced to sodium-2,2-dimethyl-2-silapentane-5-sulfonate (DSS, Cambridge Isotope Laboratories) via the water ¹H resonance and processed in MestreNova (Mestrelab Research) using 10 Hz exponential line broadening and zero filling to 262144 points.

¹³C NMR spectra were acquired at 25 °C using a Bruker Avance III HD 500 MHz instrument equipped with a BBFO+ SmartProbe. ¹³C NMR experiments, both with and without ¹H decoupling, utilized a 200 ms recycle time and a 100 ms acquisition time. All ¹³C NMR spectra were indirectly referenced to DSS (Cambridge Isotope Laboratories) via the water ¹H resonance and processed in MestreNova (Mestrelab Research) using 20 Hz exponential line broadening and zero filling to 4096 points.

¹H−¹³C heteronuclear multiple quantum coherence (HMQC) spectra were acquired on a Varian Unity Inova 500 MHz instrument equipped with an inverse triple-resonance probe. ¹H−¹³C HMQC experiments utilized a recycle time of 600 ms and a refocusing time of 2.5 ms (J_CH = 200 Hz). All ¹H−¹³C HMQC data were processed using NMRPipe with polynomial solvent correction of the time-domain data.²⁹ Both dimensions of the ¹H−¹³C HMQC data, which contained 2048 data points in the F2 dimension and 512 data points in the F1 dimension, were multiplied by a pure cosine-squared apodization function, zero filled, and Fourier transformed to obtain 1024 × 2048 real data points, after which the data were visualized in MestreNova (Mestrelab Research).

QM/MM Computations.

The initial structural model for the non-hydrogen atoms of cyanide-inhibited IsdI (IsdI−heme−CN) was derived from chain A of the X-ray crystal structure of this species (PDB ID 3QGP).¹⁰ Similarly, a structural model for all non-hydrogen atoms of N7A cyanide-inhibited IsdG (IsdG−heme−CN), with the exception of the distal cyanide ligand, was extracted from chain A of the N7A IsdG−heme crystal structure (PDB ID 2ZDO).⁹ The distal cyanide ligand was added to the N7A IsdG−heme structure using Avogadro, resulting in the initial structural model for the nonhydrogen atoms of N7A IsdG−heme−CN.^30,31 Finally, an initial structural model for the non-hydrogen atoms of IsdG−heme−CN was prepared by in silico mutation of the N7A IsdG−heme−CN structure using the Swiss-PdbViewer.³² Protein hydrogens and a 3 Å thick water sphere were added to each initial structural model using the PDB2ADF subprogram of the Amsterdam Density Functional (ADF) software package (SCM).³³⁻³⁵ This approach yield initial structural models with between 10000 and 12000 atoms.

Full QM/MM geometry optimizations of IsdI−heme−CN, WT IsdG−heme−CN, and N7A IsdG−heme−CN were completed using the ADF software package (SCM) running on the IBM Bluemoon cluster at the Vermont Advanced Computing Core.³³⁻³⁵ The QM region contained parts of the Asn7/Ala7, His77, heme, and cyanide residues, with the QM/MM boundary defined at the C_α−NH and C_α−CO bonds of the amino acids and the H₂C−CH₂COO⁻ bonds of the heme propionate side chains (Figure S3). The PBE density functional and the small core TZP basis set,^36,37 with an accint parameter of 5.0, were used for the QM region, and the AMBER95 force field was used for the MM region.³⁸ The AddRemove QM/MM coupling scheme was used, and the point charges on the QM atoms were updated each optimization cycle using simple electrostatic coupling.^39,40 The QM region was considered converged when the total energy change was less than 0.001 E_h, the maximum change in nuclear gradient was less than 0.001 E_h Å⁻¹, and the maximum change in Cartesian coordinates was less than 0.1 Å. The MM region was optimized using the conjugate gradient method and considered converged when the change in the energy gradient was less than 0.01 kcal mol⁻¹ Å⁻¹. Cartesian coordinates for all computational structures are available upon request.

The root-mean-square deviation (RMSD) between the experimental (PDB ID 3QGP) and QM/MM-optimized structures of IsdI−heme−CN, as well as between the QM/MM-optimized structures of WT and N7A IsdG−heme−CN,¹⁰ was calculated using the structural alignment of multiple proteins (STAMP) within the Visual Molecular Dynamics (VMD) program.^41,42 The secondary structures of the IsdI−heme−CN X-ray crystal structure and all three QM/MM-optimized structures were analyzed using the define secondary structure of proteins (DSSP) algorithm.^43,44 Finally, the normal-coordinate structural decomposition (NSD) method was used to quantify out-of-plane distortions of heme for all four structures noted above.^45,46

RESULTS

UV/Vis Abs Spectroscopy.

A new recombinant form of IsdG was prepared in order to increase protein stability and facilitate NMR spectroscopic characterization. The previously characterized long-linker construct of IsdG, which contains a 12 amino acid linker between the native polypeptide and the TEV protease site,⁸ polymerizes after storage at 4 °C for 1 week (Figure S4). The new short-linker construct, which contains a 2 amino acid linker, is stable at 4 °C for 1 week as assessed by SDS-PAGE gel electrophoresis (Figure S5). On the basis of CD characterization, the secondary structures of long-and short-linker IsdG are nearly identical (Figure S6). Also, on the basis of UV/vis Abs characterization, the IsdG−heme−N₃ electronic structures of the long- and short-linker constructs are the same. In both cases, the IsdG−heme−N₃ Q-band is observed at 569 nm and the Soret band blue-shifts from 418 to 417 nm upon substitution of Ala for Asn7 (Figure 2).¹² Thus, the stability of IsdG is increased by truncation of the N-terminal linker without altering the geometric or electronic structure of IsdG−heme−N₃.

Figure 2.

UV/vis Abs detected titrations of azide into WT IsdG−heme (top) and N7A IsdG−heme (bottom) in 50 mM Tris pH 7.4 and 150 mM NaCl. The traces represent samples with 0 mM (solid red), 200 mM (solid purple), and intermediate concentrations (dashed black) of azide. The UV/vis Abs detected titrations were fit to eq 1 yielding K_d values of 0.6 ± 0.1 and 2.2 ± 0.1 mM for WT and N7A IsdG−heme−N₃, respectively. The N7A substitution decreases IsdG−heme azide affinity 3-fold.

The azide dissociation constants for WT and N7A IsdG−heme−N₃ were measured in order to assess the strength of the reported hydrogen bond between Asn7 and N₃.¹² The WT IsdG−heme UV/vis Abs intensity at 425 nm was monitored as a function of azide concentration and fit to eq 1, yielding a K_d value of 0.6 ± 0.1 mM (Figure 2). Similarly, the A₄₂₅ value for N7A IsdG−heme was monitored and analyzed, resulting in a K_d value of 2.2 ± 0.1 mM for N7A IsdG−heme−N₃. This K_d increase from WT to N7A IsdG−heme−N₃ corresponds to a free energy change of 0.8 ± 0.1 kcal/mol, consistent with loss of a weak NH···N hydrogen bond upon replacement of Asn7 by Ala.^47,48 Oxygen lone pairs are also hydrogen bond acceptors; thus, it is reasonable to expect that a hydrogen bond between Asn7 and a distal oxygen-based ligand exists. Indeed, this expectation is bolstered by the previously reported observation that the UV/vis Abs spectrum of WT IsdG−heme corresponds to a low-spin Fe(III)−His−OH species, whereas the spectrum of N7A IsdG−heme is consistent with a high-spin Fe(III)−His−OH₂ species. Thus, a weak NH···N hydrogen bond between Asn7 and azide exists for IsdG− heme−N₃, and this hydrogen bond is likely also present for the putative ferric−peroxoheme intermediate of IsdG-catalyzed heme degradation.⁶

EPR Spectroscopy.

This hydrogen bond between Asn7 and azide in WT IsdG−heme−N₃ was expected to weaken the iron−azide bond and decrease the π-donor character of azide, resulting in stabilization of the Fe 3d_xz and 3d_yz based MOs relative to the Fe 3d_xy based MO. EPR spectroscopy was employed to assess this prediction, since the g values of low-spin ferric heme have been shown to be extremely sensitive probes for the relative energies of the Fe 3d_xy, Fe 3d_xz, and Fe 3d_yz based MOs.⁴⁹ The WT IsdG−heme−N₃ EPR spectrum was fit to two low-spin (S = 1/2) species with g = [2.95, 2.25, 1.49] and [2.82, 2.29, 1.61] (Figure 3). Satisfactory simulation of the low-spin component of the N7A IsdG−heme−N₃ spectrum also required the inclusion of two distinct species with g tensors of [3.03, 2.19, 1.40] and [2.81, 2.24, 1.72]. All four g tensors are consistent with ²E_g electronic ground states. In both variants, the contribution of high-spin heme to the EPR spectrum was minimal (Tables S5 and S6). The presence of two low-spin ²E_g species has been noted previously for other azide-bound ferric heme proteins,⁵⁰⁻⁵² where they were attributed to two distinct conformations of the azide ligand.

Figure 3.

Experimental (solid lines) and simulated (dashed lines) X-band (9.38 GHz) EPR spectra of WT and N7A IsdG−heme−N₃ in 125 mM KP_i pH 7.4 acquired with a microwave power of 63 μW, 100 kHz modulation frequency, 10 G modulation amplitude, and a time constant of 5.12 ms. Both spectra arise from at least two distinct low-spin ferric heme species (blue and violet dotted lines and brackets).

Ligand field theory based analysis of the g tensors for WT and N7A IsdG−heme−N₃ provided detailed insight into the iron electronic structure changes that arise from the N7A substitution. The rhombic component of the IsdG−heme−N₃ ligand field is virtually unchanged by the N7A substitution, but the axial component increases by 20% (Tables S7 and S8).²⁸ In terms of orbital energies, the unchanged rhombic component of the ligand field means that the N7A substitution does not significantly alter the relative energies of the Fe 3d_xz and 3d_yz based MOs. The 20% increase of the axial component could have several interpretations: increased spin−orbit coupling, stabilization of the Fe 3d_xy based MO, or destabilization of the Fe 3d_xz and 3d_yz based MOs. A change to spin−orbit coupling is unlikely, since the rhombic component of the ligand field is unaffected by the N7A substitution. Of the remaining options, destabilization of the Fe 3d_xz and 3d_yz based MOs is more likely since these orbitals are antibonding with respect to the azide π orbitals. As a peroxo ligand is also a hydrogen bond acceptor and π donor, it is anticipated that the putative hydrogen bond between Asn7 and the peroxo ligand in WT enzyme will perturb the relative energies of the occupied Fe 3d based MOs of ferric−peroxoheme in a similar manner.

NMR Spectroscopy.

After using EPR spectroscopy to establish that the hydrogen bond between Asn7 and azide in WT IsdG−heme−N₃ perturbs the Fe 3d based orbitals, we employed ¹H NMR spectroscopy to assess whether this electronic structure change triggered spin density delocalization onto the heme meso carbons. Previous studies have established that azide-bound heme with either high-spin ferric iron or ²E_g low-spin iron has a propensity to delocalize spin density onto the β-pyrrole carbons, resulting in downfield hyperfine-shifted heme methyl ¹H resonances in the 40−55 or 0−30 ppm range,^25,53 respectively. On the other hand, ²B_2g low-spin ferric heme tends to delocalize spin density onto the heme meso carbons, triggering upfield hyperfine shifts of the heme meso ¹H resonances.⁴⁹ As expected on the basis of the EPR data, no significantly hyperfine shifted ¹H resonances were observed in the ¹H NMR spectrum of WT IsdG−heme−N₃ (Figure S7). Thus, if WT IsdG−heme−N₃ is a thermal mixture of high-spin and ²E_g low-spin ferric heme as is the case for Mb−heme−N₃,⁵⁴ these data suggest that the equilibrium strongly favors low-spin heme for WT IsdG−heme−N₃.⁵⁵ It is also worth noting that the ¹H NMR spectrum of WT IsdG−heme−N₃ is similar to that of hydroxide-bound IsdI, where the unusual spectrum was attributed to a small amount of spin density on the heme methyl groups perhaps due to a dynamic exchange between ²E_g and ²B_2g electronic states (Figure 4).¹⁰ In contrast to WT enzyme, several ¹H resonances were observed in the 10−30 ppm region of the N7A IsdG−heme−N₃ ¹H NMR spectrum, which is consistent with the predominance of a ²E_g ground state for this species. Nevertheless, the ¹H NMR data are ambiguous regarding the question as to whether the hydrogen bond between Asn7 and azide in WT enzyme triggers spin density delocalization onto the heme meso carbons.

Figure 4.

¹H NMR spectra (25 °C) of WT (black line) and N7A (red line) IsdG−heme−N₃ in 20 mM NaP_i pH 7.4, 100 mM azide, and 10% D₂O (v/v). ¹H resonances consistent with either a ²E_g or ²B_2g ground state are not observed for WT enzyme. ¹H resonances consistent with a ²E_g ground state are observed for N7A IsdG−heme−N₃. These data strongly suggest that the N7A substitution fundamentally changes the IsdG−heme−N₃ spin density distribution.

In order to address this ambiguity, ¹³C NMR spectroscopy was employed to assess the spin density distributions in WT and N7A IsdG−heme−N₃. On the basis of the details known regarding the biosynthesis of ¹³C-heme,¹⁴ eight ¹³C resonances corresponding to the four heme meso carbons and four of the eight α-pyrrole carbons were expected in the ¹³C NMR spectrum of WT IsdG−¹³C-heme−N₃, but only seven resonances were observed (Figure 5). Integration of the spectrum revealed that the resonance at −5.8 ppm arises from two ¹³C nuclei, which accounts for the eighth ¹³C resonance. On the basis of their broadening in the ¹H-coupled ¹³C NMR spectrum, the resonances at 75.3, 46.6, 39.9, and 26.6 ppm were assigned to the heme meso carbons of WT IsdG−heme−N₃. In contrast to WT enzyme, all eight expected ¹³C resonances were readily observed in the ¹³C NMR spectrum of N7A IsdG−¹³C-heme−N₃. The resonances at 63.3, 42.5, 40.0, and 4.2 ppm were assigned to heme meso carbons on the basis of their broadening in the ¹H-coupled ¹³C NMR spectrum. Since the meso ¹³C resonances did not fully split into doublets in the ¹H-coupled ¹³C NMR spectra, two-dimensional NMR experiments were performed to verify the resonance assignments.

Figure 5.

¹H-decoupled (dashed black lines) and ¹H-coupled (solid red lines) ¹³C NMR spectra (25 °C) of WT (top) and N7A (bottom) IsdG−heme−N₃ in 20 mM NaP_i pH 7.4, 100 mM azide, and 10% D₂O (v/v). The heme meso (m) and four of the eight α-pyrrole (α) carbons have been ¹³C enriched. Resonance assignments were made on the basis of the differences between the ¹H-decoupled and ¹H-coupled spectra. The average heme meso ¹³C resonance decreases from 47.1 ppm in WT IsdG−heme−N₃ to 37.5 ppm in N7A IsdG−heme−N₃, which is consistent with significantly decreased spin density delocalization onto the heme meso carbons in the N7A variant.

¹H−¹³C HMQC spectroscopy was used to detect the heme meso ¹H−¹³C cross peaks in WT and N7A IsdG−¹³C-heme−N₃.^{56 1}H−¹³C cross peaks were observed in the WT IsdG−¹³C-heme−N₃ HMQC spectrum for the 26.6, 46.6, and 75.3 ppm ¹³C resonances (Figure 6), confirming that these three resonances arise from heme meso carbons. Similarly, three cross peaks corresponding to the 4.2, 40.0, and 63.3 ppm ¹³C resonances of N7A IsdG−¹³C-heme−N₃ were observed in the HMQC spectrum of this species. On the basis of the four heme meso 13C chemical shifts and three ¹H chemical shifts, measured for each variant, it seems reasonable to assume that the fourth heme meso carbon cross peak is buried under the residual ¹H₂O signal in the ¹H−¹³C HMQC spectra of WTànd N7A IsdG−¹³C-heme−N₃. Thus, the ¹H−¹³C HMQC data do support the resonance assignments initially made on the basis of ¹³C NMR. In order to minimize the contribution of the azide ligand orientation, all four heme meso ¹³C shifts were averaged, yielding average shifts of 47.1 and 37.5 ppm for WT and N7A IsdG−heme−N₃, respectively. The downfield shift of the average heme meso ¹³C shift in WT enzyme in comparison to the N7A variant is consistent with a significant increase in heme meso carbon spin density upon introducing a hydrogen bond between Asn7 and azide.⁴⁹ Since a peroxo ligand is also a hydrogen bond accepting π-donor ligand, it is expected that the putative hydrogen bond between Asn7 and peroxo also increases heme meso carbon spin density in ferric−peroxoheme.

Figure 6.

¹H−¹³C HMQC spectra (25 °C) of WT (black) and N7A (red) IsdG−heme−N₃ in 20 mM NaP_i pH 7.4, 100 mM azide, and 10% D₂O (v/v). For each variant, three of the four expected heme meso ¹H−¹³C cross peaks (m) are detectable. These data provide further support for the ¹³C resonance assignments presented in Figure 5.

QM/MM Computations.

Previous studies have established that porphyrin ruffling can trigger spin density delocalization onto the heme meso carbons in noncanonical heme oxygenases;^10,11,46,57 therefore, QM/MM calculations were employed to assess whether the electronic structure changes detected by ¹³C NMR could arise from ruffling of the substrate by Asn7. Since QM/MM modeling of noncanonical heme oxygenases has not been reported previously, the computational setup was initially benchmarked using the X-ray crystal structure of IsdI−heme−CN (PDB ID 3QGP). On the basis of the 0.7 Å RMSD between the C_α atoms of the QM/MM and crystallographic models of IsdI−heme−CN (Table S9), the hybrid PBE and AMBER95 model employed here accurately models the tertiary structure of noncanonical heme oxygenases.^36,38,41 DSSP analysis of the computational model and X-ray crystal structure identified 2% and 3% increases in the α-helix and β-sheet contributions, respectively, to the QM/MM secondary structure in comparison to the crystallographic structure (Table S10).^43,44 These minor secondary structural differences may arise from the fact that the QM/MM model is an aqueous monomer, but this form of the protein crystallized as a dimer. Finally, the QM/MM calculations do an excellent job modeling the ruffling distortion of heme. NSD analysis of the computational structure revealed a 2.24 Å ruffling distortion for the heme substrate,⁴⁵ which differs from the crystallographic value of 2.25 Å by 0.01 Å. On the basis of these analyses, the choices of QM model, MM model, and QM/MM boundary used here are adequate to produce accurate structural models of noncanonical heme oxygenases.

Once the QM/MM setup was benchmarked for IsdI−heme−CN, a QM/MM optimization of N7A IsdG−heme−CN was completed by starting from the X-ray crystal structure of N7A IsdG−heme (PDB ID 2ZDO).⁹ Cyanide was used as a distal ligand in these calculations because this was the coordination complex moiety benchmarked above and there are currently no X-ray crystal structures of an azide-inhibited noncanonical heme oxygenase available to benchmark an IsdG−heme−N₃ QM/MM model. On the basis of the reported differences between the electronic structures of IsdG−heme−CN and IsdG−heme−N₃,¹² it is important to limit analyses of these computational data to geometric, and not electronic, structure. It should also be noted out that the QM/MM-optimized structures do not capture protein dynamics, and NMR studies have elucidated important differences between the microsecond to millisecond time scale motions of azide- and cyanide-inhibited canonical heme oxygenases.⁵⁸ With these caveats in place, it can be reported that the RMSD for the C_α atoms for X-ray crystal structure of N7A IsdG−heme (PDB ID 2ZDO) and the N7A IsdG−heme−CN QM/MM structure is 0.8 Å (Table S9). DSSP analysis revealed that these models have similar secondary structures,^43,44 with the QM/MM model exhibiting a 1% decrease in α-helices and a 2% increase in β-sheets (Table S11). Finally, these data suggest that cyanide binding only increases the magnitude of heme ruffling by 0.13 Å from 1.88 to 2.01 Å in N7A IsdG−heme. Thus, cyanide binding does not significantly alter the minimum energy structure of IsdG−heme. As a consequence, it is expected that azide binding also will not significantly perturb the IsdG−heme geometry.

Finally, a QM/MM optimization of WT IsdG−heme−CN was performed by starting from the X-ray crystal structure of N7A IsdG−heme (PDB ID 2ZDO) to investigate the influence of Asn7 on heme ruffling.⁹ An X-ray crystal structure of heme-bound WT IsdG is not currently available; therefore, the Asn7 residue was introduced into the N7A IsdG−heme crystal structure in silico. The C_α atoms of the WT and N7A IsdG−heme−CN models have an RMSD value of 0.4 Å, in accord with the similarities between the X-ray crystal structures of WT IsdG (PDB ID 1XBW) and N7A IsdG−heme (PDB ID 2ZDO, Figure 7 and Table S9).^8,9 Furthermore, DSSP analysis revealed that the secondary structures of WT and N7A IsdG−heme−CN are nearly identical, with only a 1% increase in β-sheet content upon making the N7A substitution (Table S11).^43,44 Thus, it is perhaps not surprising that the N7A substitution only decreases the heme ruffling contribution by 0.03 Å from 2.04 to 2.01 Å according to an NSD analysis of the QM/MM structures.^45,46 It is important to note that this 0.03 Å change is an order of magnitude smaller than the 0.4−0.8 Å decrease reported for W66Y IsdI−heme−CN, which results in an electronic ground state change that alters heme meso spin density.¹¹ In summary, the QM/MM data reported here strongly argue against the decreased heme meso spin density in N7A IsdG−heme−N₃ being derived from a decrease in ruffling and indirectly support the conclusion that a hydrogen bond between Asn7 and the distal ligand tunes the electronic structure of porphyrin.

Figure 7.

Structural alignment of the PBE/TZP + AMBER95 QM/MM-optimized structures of WT (light green) and N7A (light blue) IsdG−heme−CN. The PBE/TZP region included parts of the Asn7/Ala7, His77, heme, and cyanide residues. The N7A substitution does not significantly influence the magnitude of heme ruffling within the IsdG active site.

DISCUSSION

Electronic Structure of the Ferric−Peroxoheme Intermediate.

The multifaceted characterization of IsdG−heme−N₃ presented in this article yielded detailed insight into the influence of Asn7 on the electronic structure of this species. Previous work has shown that heme ruffling by noncanonical heme oxygenase active sites tunes the substrate electronic structure and reactivity,^{6,10,11,21,46,57} but benchmarked QM/MM calculations presented here demonstrated that Asn7 does not contribute to porphyrin ruffling. Instead, the spectroscopic data presented here indicate that there is long-range electronic communication between Asn7 and the porphyrin meso carbons via azide and iron. UV/vis Abs detected titrations of azide into WT and N7A IsdG−heme strongly suggest a hydrogen bond between Asn7 and azide. EPR spectroscopic data reveal that this hydrogen bond tunes the iron electronic structure by stabilizing the Fe 3d_xz and 3d_yz based MOs without changing the electronic ground state. Finally, ¹³C NMR data for WT and N7A IsdG−heme−N₃ demonstrate that this iron electronic structure change delocalizes spin density onto the heme meso carbons, presumably due to mixing of the Fe 3d_xy and porphyrin a_2u orbitals.⁴⁹ Altogether, these data provide the first unequivocal evidence for long-range electronic communication between the conserved second-sphere Asn residue and the porphyrin meso carbons in noncanonical heme oxygenases.

These data provide important insight into the electronic structure of the critical ferric−peroxoheme intermediate, because azide is an excellent analogue for peroxide. Previous studies of noncanonical heme oxygenases have employed cyanide as an analogue for peroxide, ^{10−12,21,46,57} but cyanide is a π*-acceptor ligand whereas peroxide is a π-donor ligand and the two are expected to have fundamentally different influences on the electronic structure of iron. One option for a π-donor analogue of peroxide is hydroxide, but this ligand is in equilibrium with a water ligand at physiological pH. On the other hand, azide is a π-donor ligand that is not readily protonated at pH 7.4. Furthermore, the π-donor strengths of azide and peroxide are expected to be similar because the π-nonbonding HOMO of azide and the π*-antibonding HOMO of peroxide are expected to be at similar energies due to the increased electronegativity of oxygen relative to nitrogen. Another important similarity between azide and peroxide is that both can accept a hydrogen bond via their iron-ligating α-atom. Considering all of these similarities between azide and peroxide, the data presented in this article strongly suggest that a hydrogen bond between Asn7 and peroxide triggers spin density delocalization onto the heme meso carbons in the critical ferric−peroxoheme intermediate of IsdG without changing the electronic ground state (Figure 8).

Figure 8.

Spectroscopic data presented in this article strongly suggest that a hydrogen bond between Asn7 and the peroxo ligand stabilizes the Fe 3d_xz and 3d_yz based molecular orbitals without changing the electronic ground state (top). This iron electronic structure change alters spin density (gray dots) of the porphyrin ligand, resulting in meso carbons with electrophilic radical character (bottom).

Origin of Novel Heme−Oxygen Chemistry.

The electronic structure data presented above provide detailed insight into the electronic contributions to IsdG-catalyzed heme degradation. Previous studies have revealed the structural contributions of the enzyme active site to porphyrin oxygenation; the terminal oxygen of the peroxo ligand is bent toward the β- or δ-meso carbon by Asn7,¹² and these two carbons are pushed up out of the heme plane toward the peroxo ligand by Trp67.^10,11 This study provides the first direct experimental support for a recently hypothesized electronic contribution of Asn7 to heme oxygenation; a hydrogen bond between Asn7 and peroxide increases heme meso carbon spin density without changing the electronic ground state, resulting in downfield hyperfine-shifted heme meso ¹³C resonances. This electronic structure change can be thought of as a larger contribution from a ferrous porphyrin cation radical resonance structure where the unpaired electron is delocalized among the four porphyrin meso carbons.⁵⁹ This resonance structure promotes a rearrangement process with concerted O−O bond cleavage and C−O bond formation by increasing the electrophilicity and partial radical character of the heme meso carbons. The activation energy for this reaction is expected to be lower than the 12.6 kcal/mol barrier predicted previously by a computational model that neglected the contribution of Asn7.⁶ The nearly complete picture of IsdG-catalyzed heme monooxygenation that emerges from these data is that Asn7 and Trp67 work in concert to promote regioselective porphyrin oxygenation by exercising structural and electronic control over the substrate.

The realization that both structural and electronic contributions from both Asn7 and Trp67 are required for the unique rearrangement oxygenation explains why this chemistry has not been observed in other heme proteins. A steric interaction with Trp67 is responsible for pushing the β-and δ-meso carbons above the porphyrin plane toward the oxygen binding site,^10,11 and a hydrogen bond with Asn7 is responsible for rotating the distal peroxo ligand toward the β-or δ-meso carbon,¹² which brings both components of the new C−O bond in close proximity (Figure 9). The b_1u ruffling deformation induced by Trp67 changes the point group of the heme moiety to D_2d, turning on Fe 3d_xy and porphyrin a_2u orbital mixing, ^45,49the Asn7 hydrogen bond perturbs the Fe(III) electronic structure, resulting in an increased Fe 3d_xy contribution to the singly occupied molecular orbital, and the combination results in electrophilic porphyrin meso carbons with partial radical character. Canonical heme oxygenases and monoheme peroxidases do not ruffle heme;¹ therefore, these enzymes do not bring a heme meso carbon and the terminal oxygen of a peroxo ligand into close proximity nor do they turn on Fe 3d_xy and porphyrin a_2u mixing. On the other hand, nitrophorins induce a significant ruffling deformation but lack a hydrogen bond donor in the distal pocket;⁶⁰ thus, these enzymes cannot induce the electronic structure change required for IsdG-like heme oxygenation. Interestingly, MhuD heme degrading enzymes have conserved residues equivalent to Trp67 and Asn7,⁶¹ and the arrangement of these residues relative to the heme substrate should result in the formation of a mixture of α- and γ-mesohydroxyheme by following an IsdG-like mechanism.⁵⁷ However, recent data strongly suggest that WT MhuD generates β- and δ-mesohydroxyheme en route to mycobilin production⁶² and R26S MhuD catalyzes formation of biliverdin;⁶³ therefore, MhuD must have a distinct heme oxygenation mechanism in comparison to IsdG. As a result of this observation, the role of Asn7 in MhuD-catalyzed heme oxygenation is under investigation by our laboratory.

Figure 9.

Data presented in this article have revealed how Asn7 and Trp67 work in concert to promote IsdG-catalyzed heme hydroxylation (top). A steric interaction between Trp67 and the porphyrin ligand pushes the β- and δ-meso carbons toward the peroxo ligand and turns on mixing of the Fe 3d_xy and porphyrin a_2u orbitals (middle). A hydrogen bond between Asn7 and the peroxo ligand orients the bent peroxo ligand along the β-/δ-meso carbon axis and increases the contribution of the Fe 3d_xy orbital to the singly occupied molecular orbital of the complex (bottom). The combination of these two perturbations results in spin density delocalization onto the heme meso carbons and an electronic driving force for heme hydroxylation. Similar hydroxylation chemistry is not observed in canonical heme oxygenases, peroxidases, or nitrophorins because both second-sphere interactions are required to promote the rearrangement reaction catalyzed by IsdG.

CONCLUSION

A detailed spectroscopic and computational investigation of IsdG has been completed that provides valuable insight into the mechanism of the first heme oxygenation reaction catalyzed by this enzyme. Spectroscopic characterizations of WT and N7A IsdG−heme−N₃ revealed that a heme electronic structure change introduced by the second-sphere Asn7 residue results in spin density delocalization onto the porphyrin meso carbons without an electronic ground state change from ²E_g to ²B_2g. Computational modeling confirmed that this electronic structure change arose from an Asn···N₃ hydrogen bond, and not from an alteration of heme ruffling. Due to the important similarities between the metal coordination properties of azide and peroxide, these data strongly suggest that partial porphyrin to iron electron transfer in the critical ferric−peroxoheme intermediate of IsdG results in heme meso carbons with partial cation radical character. This electronic structure change only arises from concerted perturbations of heme by second-sphere Asn and Trp residues, which explains why IsdG-like chemistry is not commonly observed in His-ligated heme proteins.