Nitrosyl Myoglobins and Their Nitrite Precursors: Crystal Structural and QM/MM Theoretical Investigations into Preferred Fe–NO Ligand Orientations in Myoglobin Distal Pockets

Abstract

The globular dioxygen-binding heme protein myoglobin (Mb) is present in several species. Its interactions with the simple nitrogen oxides, namely nitric oxide (NO) and nitrite, have been known for decades, but the physiological relevance has only recently become more fully appreciated. We previously reported the O-nitrito binding mode of nitrite to ferric horse heart wild-type (wt) Mb^III and human hemoglobin. We have expanded on this work and report the interactions of nitrite with wt sperm whale (sw) Mb^III and its H64A, H64Q and V68A/I107Y mutants whose dissociation constants increase in the order H64Q < wt < V68A/I107Y < H64A. We also report their X-ray crystal structures that reveal the O-nitrito binding mode of nitrite to these derivatives. The Mb^II-mediated reductions of nitrite to NO and structural data for the wt and mutant Mb^II–NOs are described. We show that their FeNO orientations vary with distal pocket identity, with the FeNO moieties pointing towards the hydrophobic interiors when the His64 residue is present, but pointing towards the hydrophilic exterior in the absence of this His64 residue. This correlates with the nature of H-bonding to the bound NO ligand (nitrosyl O vs. N atom). Quantum mechanics and hybrid quantum mechanics/molecular mechanics calculations help elucidate the origin of the experimentally preferred NO orientations. In a few cases, the calculations reproduce the experimentally observed orientations only when the whole protein is taken into consideration.

Table of Contents Graphic

INTRODUCTION

Since its first mention as the red protein “myochrome” by Mörner in 1897,¹ the globular heme protein myoglobin (Mb) has continued to intrigue researchers across a broad range of fields.² The interaction of Mb with the simple nitrogen oxides (NOx; x = 1, 2), including the use of nitrite (NO₂-) in the meat curing industry, has a long history.^3–5 The now widely-recognized vasodilatory property of nitrite continues to attract intense interest, especially in its interactions with heme proteins.^6–8

Nitrite is an intermediate in denitrification,^9–13 and its enzymatic reduction to NO (eq. 1) is of direct relevance to the global N-cycle.^{14, 15} Much of nitrite biology involves its interactions with metals,^{15, 16} necessitating consideration of its coordination chemistry.

$${\text{NO}}_3^{-}\to {\text{NO}}_2^{-}\to \text{NO}\to {\text{N}}_{\text{2}}\text{O}\to {\text{N}}_{\text{2}}$$ (1)

Three main binding modes of nitrite to metals have been structurally characterized (Figure 1),¹⁷ although only the first two (N-nitro and O-nitrito) have been determined for heme proteins and heme models to date.

Figure 1.

The three common binding modes for nitrite to metals.

The N-nitro binding mode remains the most common binding mode of nitrite to heme Fe centers,^18–31 despite the differences in proximal ligation to heme (e.g., His for cyt cd₁ NiR, Cys for SiR, and Lys for cyt c NiR), macrocycle saturation (e.g., d vs c heme), and distal pocket amino acid identities. The first suggestions of a possible O-nitrito binding mode to a heme protein was that for ferric Hb^III–ONO by Hartridge in 1920³² and Barnard in 1937,³³ although this binding mode was not structurally confirmed until almost a century later with our reports on the structures of the nitrite adducts of horse heart (hh) Mb^III and derivatives,^34–36 and ferric human hemoglobin (Hb^III).^{37, 38}

Support for low energy differences between the N-nitro and O-nitrito binding modes in heme proteins and their truncated active site models comes from density functional theory (DFT) calculations that predict the N-nitro mode to be lower in energy by a few kcals/mol.^{19, 31, 39–46} A recent detailed quantum chemical analysis of EPR spectral data for Mb^III–nitrite,⁴⁷ however, supports the O-nitrito binding mode to the metal center as observed experimentally. These results suggest that the Mb (and Hb) distal pocket environments with the single His64 H-bonding residue in these proteins are somewhat unique in their interactions with nitrite. Indeed, we reported that although the hh wt Mb^III–nitrite complex displays the O-nitrito binding mode, its hh H64V mutant displayed a long Fe–N distance of ~2.6 Å suggestive of an electrostatic interaction with nitrite (~65% nitrite occupancy) in an apparent N-nitro form (Figure 2, left).⁴⁸ The recently obtained structure of the sperm whale (sw) Mb^III–nitrite analog (Figure 2, right) displays a more defined N-nitro mode (Fe–N = 2.1 Å) illustrating the importance of the His64 ligand (or lack thereof) in this nitrite orientation.

Figure 2.

Models and F_o–F_c omit maps showing nitrite in the distal pockets of the H64V mutants of ferric hh Mb^III–nitrite (left; PDB ID: 3HEP) and ferric sw Mb^III–nitrite (right; PDB ID: 6CF0).

The reduction of nitrite to NO in denitrification as part of the global N-cycle proceeds via the reaction shown in eq. 2 that is catalyzed by the NiR enzymes.^{9, 10, 12, 16} Haldane first

$${\text{NO}}_2^{-}+2{\text{H}}^{+}+1{\text{e}}^{-}\to \text{NO}+{\text{H}}_2\text{O}$$ (2)

suggested in 1901, based on his studies on salted meat using visible spectroscopy, that ferrous Hb^II reduced nitrite to the nitrosyl derivative Hb^II–NO,⁵ and this was later elaborated by Brooks in 1937.⁴⁹ Interestingly, ferrous Mb^IIs from several species have now been shown to participate in nitrite reduction to generate NO (eqs. 3–4).^{3, 50} Other metalloproteins such as Hb,⁵¹ endothelial NO synthase,^{52, 53} xanthine oxidase,⁵⁴ neuroglobin,^{55, 56} cytoglobin,^57–59 and some plant Hbs⁶⁰ similarly reduce nitrite to NO.

$${\text{Mb}}^{\text{II}}+{\text{NO}}_2^{-}+{\text{H}}^{+}\to \text{NO}+{\text{Mb}}^{\text{III}}+{\text{OH}}^{-}$$ (3)

$${\text{Mb}}^{\text{II}}+\text{NO}\to {\text{Mb}}^{\text{II\_}}\text{NO}$$ (4)

It is somewhat surprising that although MbNO and HbNO have biological relevance, not much is currently known about the FeNO conformational preferences in these proteins as a function of varied distal pocket H-bonding environments. Indeed, only a small handful of Mb^II–NO and Hb^II–NO crystal structures have been reported to date.^{61,34, 62,63,64, 65} It is important to note that the NO vectors point towards the hydrophobic interiors in all these Mb^II–NO and Hb^II–NO structures reported to date.⁶⁶ Given that Mbs from several species reduce nitrite to generate their Mb^II–NO derivatives, we were interested in determining the effects of distal pocket mutations on both the binding modes of the parent Mb-nitrite adducts and the orientations of the Fe–NO moieties in their Mb^II–NO derivatives. In particular, we note that although distal pocket H-bonding to the nitrosyl N-atom has been demonstrated in Mb^II–NO, the corresponding (sole) H-bonding to the nitrosyl O-atom has not yet been reported.

In this paper, we report the preparation and crystal structures of the nitrite and NO complexes of wt and mutant sw Mbs, with the goal of examining the nature of nitrite binding (N-nitro vs. O-nitrito) and determining the FeNO orientation preferences in these wt and mutant Mbs. We chose the mutants shown in Figure 3 that differ in H-bonding capacity to the bound NOx ligands (wt with 1 H-bonding residue; H64A with 0 H-bonding residues; H64Q with 1 H- bonding residue; V68A/I107Y with 2 H-bonding residues), which to the best our knowledge, is the first systematic structural investigation of different H-bonding capacities on NO orientations in the heme distal pocket, including some new patterns not reported previously. The H64Q mutant is an analog of the wt Mbs from certain species of worms,^{67, 68} elephants,^{69, 70} and sharks,^71–73 and of the α chains of opossum Hbs.⁷⁴ Alternatively, the H64A mutant is related to the Mbs that have hydrophobic residues (Val and Leu) at this E7 position.^75–78 We hypothesized that the introduction of the potential second H-bond interaction in the V68A/I107Y double mutant would help stabilize the V-shaped N-nitro binding mode as observed in cyt cd₁ NiRs. In addition to the experimental characterization of the Mb^III–nitrite and Mb^II–NO products, we sought to probe the factors that influence the FeNO H-bonding and orientational preferences in the wt and mutant Mb^II–NOs by quantum mechanics (QM) and hybrid quantum mechanics and molecular mechanics (QM/MM) calculations. Importantly, such a combined experimental and computational study shows for the first time that the FeNO orientations vary as a function of distal pocket identity.

Figure 3.

The wt and mutant Mbs used in this study.

MATERIALS AND METHODS

Expression, purification and crystallization of wild-type and mutant sperm whale myoglobins

The recombinant wt sw Mb plasmid was a kind gift from Dr. Mario Rivera (Univ. of Kansas); the recombinant wt protein contains a D122N mutation (due to an error in the former’s sequence determination)⁷⁹ and a Met residue at the N-terminus. Mutagenesis was performed using the Quick-Change method (Strategene). For the double mutant V68A/I107Y, the V68A mutation was performed first followed by the I017 mutation. The wt, H64A, H64Q and V68A/I107Y mutants were expressed in E. coli BL21(DE3) and purified as described by Springer and Sligar.⁸⁰ The mutant H64A and H64Q proteins had higher and more stable expression levels (~100 mg/mL) compared with those of the wt and double-mutant V68A/I107Y proteins (~10 mg/mL). Crystals of the recombinant ferric Mb^III proteins were grown as described previously.^{81, 82}

Reactions of the sw Mb proteins with NOx

Sodium nitrite, potassium ferricyanide, and sodium dithionite were purchased from Sigma-Aldrich and used as received.

UV-vis monitoring of the reactions:

A Hewlett Packard 8453 spectrophotometer was used to obtain the UV-vis spectra of the complexes to determine the reaction times and the stabilities of the complexes. The protein samples in 20 mM Tris, 1 mM EDTA, pH 7.4 were purified as stated above.⁸⁰ UV-vis spectra were recorded for all the protein derivatives; 10–15 μL of protein (10–20 mg/mL) were added into 2.5 mL 0.1 M phosphate buffer at pH 7.4. Next, 5 μL of 1 M sodium nitrite in phosphate buffer were added to each of the protein samples to obtain the spectra of sw Mb^III–nitrite adducts. To obtain the spectra of the sw Mb^II–NO derivatives, the Mb^III samples were reduced by the addition of 5 μL of 1.0 M dithionite, followed by the addition of 5 μL of 1.0 M sodium nitrite into the respective cuvettes. The spectra were recorded at different time points.

Determination of nitrite reductase activity of ferrous Mb^II and nitrite binding constants of Mb^III:

The Mb protein samples were fully oxidized by the addition of excess of potassium ferricyanide,⁸³ and the mixtures passed through Sephadex G25 columns to remove the excess reagent. The protein fractions were concentrated to ~20 mg/mL. The measurement of nitrite reductase activity of ferrous Mb^IIs and the dissociation constant K_D for Mbs followed published protocols.^{55, 84}

Preparation of the crystalline derivatives:

Crystals of the Mb–NOx complexes were obtained by the soaking method using crystals of the respective Mb^IIIs obtained by the method of Phillips et al.⁷⁹ Prior to reactions with the NOx, a suitably sized crystal of the ferric Mb^III–H₂O precursor was harvested for X-ray diffraction data collection and its crystal structure solved. The structures of the deoxy Mb^IIs without any ligands were solved using crystals derived from the dithionite reduction of the Mb^III–H₂O precursors. Specifically, appropriately sized Mb^III–H₂O crystals were transferred into a 4 μL droplet (well solution containing 100 mM Tris, 1 mM EDTA, pH 7.4, 2.6 M (NH₄)₂SO₄, with 10% glycerol) on a cover slide. The cover slide with the drop was placed in a layer of light mineral oil to minimize the introduction of air. A few crystals of solid dithionite were added into the droplet, and the color changed from red to pink during the ensuing ~10 min period, signifying the formation of the deoxyMb^II. The treated crystals were harvested using cryo-loops and frozen in liquid nitrogen prior to X-ray diffraction data collection.

The Mb^III–nitrite complexes were obtained using a similar method as described above. A few solid crystalline particles of sodium nitrite were added into the droplet to react with the respective Mb^III–H₂O crystals for 20–30 min. The treated crystals were harvested using cryo-loops at different time periods and frozen in liquid nitrogen prior to X-ray diffraction data collection.

Crystals of the ferrous Mb^II–NO derivatives were obtained by soaking the respective sw Mb^III–H₂O crystals with nitrite for 20–30 min, followed by the addition of excess solid dithionite to reduce the Mb^III–nitrite intermediate. In these reactions, a clear color change from brown to pink was observed. The treated crystals were harvested as above.

X-ray diffraction data collection, processing, structure solution, and refinement

The diffraction data were collected at home source using a Rigaku RU-H3R and/or MicroMax 007HF microfocus X-ray generator, coupled to a R-AXIS IV⁺⁺ detector or a PILATUS 200K detector. The data were collected at 100 K with CuKα radiation (λ = 1.54178 Å) from the generator operated at 50 kV/100 mA (Rigaku RU-H3R generator) or 40 kV/30 mA (Rigaku MicroMax 007HF generator).

The data were processed using iMOSFLM^{85, 86} or HKL3000.⁸⁷ The output mtz files from iMOSFLM were scaled by Aimless or Scala.⁸⁸ When HKL3000 was used, the sca output files were transformed into mtz files by Scalepack2mtz as implemented in the CCP4 program suite.⁸⁹

Initial phases were obtained by molecular replacement using PHASER (CCP4).⁹⁰ The starting model used was wt sw Mb^III–H₂O at 1.5 Å resolution (PDB ID: 2MBW) with the heme, sulfates, and water molecules removed from the structure. Refinements were performed using either Refmac5 (CCP4) or Phenix.refine (PHENIX).^{91, 92} Models were rebuilt using COOT.⁹³ The MolProbity server was used for structure validation.⁹⁴ The details are described below for each of the structures.

The figures were generated using PYMOl.⁹⁵ 2F_o–F_c electron density maps were initially calculated by Fast Fourier Transform (FFT) in the CCP4 software package.⁹⁶ The resulting map files were converted to ccp4 files and displayed in PYMOL. In general, each F_o–F_c electron density map was generated as follows: (i) a new PDB file in which ligands were removed from the pocket was initially refined for 5 cycles in Refmac5; (ii) the new mtz file that was generated from the first step was input into FFT to generate the map that was then displayed in PYMOL.

A note on previously published crystal structures:

Sixteen sw Mb crystal structures are reported in this paper. Five reference (Mb^III, deoxyMb^II) structures used in this work have been reported previously: wt Mb^III–H₂O,⁹⁷ wt deoxyMb^II,⁹⁷ H64A Mb^III–H₂O (PDB ID: 102M),⁸² H64Q Mb^III–H₂O (PDB ID: 2MGH),⁷⁹ and H64Q deoxyMb^II (PDB ID: 2MGG).⁷⁹ We reproduced these five structures for a more realistic comparison (see Supporting Information) with our new structures detailed below. Only one sw Mb–NOx structure has been reported to date, namely sw wt Mb^II–NO (PDB ID: 1HJT) prepared from gaseous NO addition (after sodium fluoride addition) to ferric Mb^III.⁶¹ The new structures reported in this work are described below.

wt Mb^III–nitrite:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.470 to 0.318. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. Three sulfate anions, seven nitrite anions and one glycerol were added iteratively using COOT. Two conformations for each of the sidechains of Lys133 and Lys145 were modeled with 50% occupancy each. The final model was refined to 1.85 Å resolution with an R factor of 0.148 and R_free of 0.187.

wt Mb^II–NO:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.444 to 0.307. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. One sulfate anion, one nitrite anion, one nitric oxide and one glycerol molecule were added iteratively using COOT. The C-terminal residues Gln152 and Gly153 were omitted due to the lack of electron density. The final model was refined to 1.70 Å resolution with an R factor of 0.138 and R_free of 0.191.

H64A deoxyMb^II:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.372 to 0.255. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. Six sulfate anions and one glycerol molecule were added iteratively using COOT. Two conformations for each of the sidechains of Leu115, Asn132 and Tyr151 were modeled with 50% occupancy each. The N-terminal Met residue was omitted due to the lack of electron density. The final model was refined to 1.78 Å resolution with an R factor of 0.148 and R_free of 0.171.

H64A Mb^III–nitrite:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.347 to 0.255. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. Three sulfate anions, eight nitrite anions and one glycerol molecule were added iteratively using COOT. Two conformations for each of the sidechains of Asn122 were modeled with 50% occupancy each. The N-terminal Met residue was omitted due to the lack of electron density. The final model was refined to 1.85 Å resolution with an R factor of 0.148 and R_free of 0.187.

H64A Mb^II–NO:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.334 to 0.245. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. Three sulfate anions, seven nitrite anions, one nitric oxide and two glycerol molecules were added iteratively using COOT. Two conformations for each of the sidechains of His12, Asn122, and Lys133 were modeled with 50% occupancy each. The N-terminal Met residue was omitted due to the lack of electron density. The final model was refined to 1.78 Å resolution with an R factor of 0.149 and R_free of 0.186.

H64Q Mb^III–nitrite:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.294 to 0.222. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. Two sulfate anions, ten nitrite anions and two glycerol molecules were added iteratively using COOT. Two conformations for each of the sidechains of Lys16, Val68 and Asn122 were modeled with 50% occupancy each. The N-terminal Met residue was omitted due to the lack of electron density. The final model was refined to 1.81 Å resolution with an R factor of 0.160 and R_free of 0.201.

H64Q Mb^II–NO:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.317 to 0.240. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. Three sulfate anions, seven nitrite anions, one nitric oxide and three glycerol molecules were added iteratively using COOT. Two conformations for each of the sidechains of Asn122 were modeled with 50% occupancy each. The N-terminal Met residue was omitted due to the lack of electron density. The final model was refined to 1.78 Å resolution with an R factor of 0.153 and R_free of 0.193.

V68A/I107Y Mb^III–H₂O:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.371 to 0.240. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. Three sulfate anions and one glycerol molecule were added iteratively using COOT. The C-terminal residues Gln152 and Gly153 were omitted due to the lack of electron density. The final model was refined to 1.78 Å resolution with an R factor of 0.160 and R_free of 0.199.

V68A/I107Y deoxyMb^II:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.384 to 0.244. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. Two sulfate anions and two glycerol molecules were added iteratively using COOT. The N-terminal Met residue and the C-terminal residues Gln152 and Gly153 were omitted due to the lack of electron density. The final model was refined to 1.80 Å resolution with an R factor of 0.186 and R_free of 0.250.

V68A/I107Y Mb^III–nitrite:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.455 to 0.276. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. One sulfate anion and two nitrite anions were added iteratively using COOT. The N-terminal Met residue and the C-terminal residues Gln152 and Gly153 were omitted due to the lack of electron density. The final model was refined to 1.57 Å resolution with an R factor of 0.175 and R_free of 0.247.

V68A/I107Y Mb^II–NO:

Ten initial cycles of restrained refinement were run with Refmac, and the R factor decreased from 0.433 to 0.282. Ligands and water were added to the model depending on the F_o–F_c electron density map in the subsequent refinement cycles. One sulfate anion, one nitrite anion and one nitric oxide were added iteratively using COOT. The N-terminal Met residue was omitted due to the lack of electron density. The final model was refined to 1.79 Å resolution with an R factor of 0.157 and R_free of 0.205.

The X-ray data collection and refinement statistics are collected in Table S1 in the Supporting Information.

A note on the reproducibility of the O-nitrito conformation:

We previously reported an apparent N-nitro to O-nitrito linkage isomerization upon exposure of chlorin-substituted MbChl(NO₂) complex to multiple cycles of home-source X-ray data collection.⁹⁸ To probe if such a linkage isomerization was occurring in the Mb^III–nitrite complexes reported here, we collected multiple X-ray diffraction data sets for each Mb^III–nitrite complex using variable total X-ray radiation exposure times. Only the O-nitrito binding mode to the heme centers was observed in each case.

Computational Details

Two sets of calculations were used to investigate the Fe^II–NO orientations in the nitrosyl derivatives of the wt and mutant Mbs: (i) quantum mechanics (QM) calculations on active site models, and (ii) hybrid quantum mechanics and molecular mechanics (QM/MM) calculations on the whole protein derivatives. All calculations were performed using the program Gaussian 09.⁹⁹

In the first set of QM calculations, in order to account for any H-bonding and van der Waals interactions involving the NO ligand, all residues with any atoms within a 7 Å distance to Fe were included in the active site models, except for the case of the H64Q Mb^II–NO compound, in which the Thr67 residue was not included to avoid the artificial H-bond formation with Gln64 in this truncated model (this H-bonding interaction was not present in the experimentally determined X-ray structure). The specific residues accounted for in the active site models are listed in Table 1. A non-substituted porphyrin macrocycle was used. The partial geometry optimizations, with Fe and Cα atoms of residues fixed during the geometry optimization to mimic the protein environment, were conducted for all the active site models based on similar studies.^{100, 101} Frequency calculations were subsequently performed to verify the nature of the corresponding stationary states on their potential energy surfaces and provide zero-point energy corrected electronic energies (E_ZPE’s), enthalpies (H’s), and Gibbs free energies (G’s). Based on a recent methodology study for similar model systems,^{102, 103} geometries were optimized using the mPW1PW91 method,¹⁰⁴ together with the self-consistent reaction field method using the PCM approach^105–108 with a dielectric constant of 4.0 to simulate the protein environment. For the geometry optimization, we utilized (i) the Wachters’ basis¹⁰⁹ for Fe, (ii) 6–311++G(2d,2p) for first shell atoms (the four coordinated porphine N atoms and the axial histidine N atom, the NO ligand, and the heavy atom forming the H-bond with NO), and (iii) 6–31G(d) for the other atoms.

Table 1.

Residues involved in optimization in the QM calculations.

Protein	Residues
wt MbII–NO	Leu29, Phe43, His64, Val68, His93, Ile107
H64A MbII–NO	Leu29, Phe43, Ala64, Thr67, Val68, His93, Ile107
H64Q MbII–NO	Leu29, Phe43, Gln64, Val68, His93, Ile107
V68A/I107Y MbII–NO	Leu29, Phe43, His64, Ala68, His93, Tyr107

The QM/MM calculations were performed using the ONIOM approach.¹¹⁰ All the protein X-ray structures were first inspected for any potential problems and edited with the solvent, NO₂^-, SO₄^2-, and glycerol molecules removed. The QM region consists of Fe, the truncated porphine macrocycle, the axial His93 and the NO ligand. In the ONIOM treatment of the QM/MM boundary atoms, His93 was cut at the C_α–C_β bond. The QM region was treated with the same method and basis set as described above for the QM calculations for the active site models. The MM region was described using the Amber force field¹¹¹ as used in typical ONIOM calculations.^112–114 The Merz-Singh-Kollman¹¹⁵ charge of NO and Amber^{111, 116} charges for other atoms were used as the MM charges. The details of treatment of a few missing Amber parameters for the current Mb^II–NO systems have been described recently in similar work for Mb^II–HNO.¹¹⁷

RESULTS AND DISCUSSION

Nitrite binding to the ferric wt and mutant sw Mb^IIIs

Heme nitrite complexes such as Mb^III–nitrite are generally prepared from adduct formation between nitrite and the ferric heme protein (eq. 5). An alternate but far less common route to their formation occurs from attack of NO on ferryl heme species,^{118, 119, 120} where the O-nitrito formulations for Mb¹¹⁸ and Hb¹¹⁹ (eq. 6) were based on UV-vis spectroscopy.

$${\text{Mb}}^{\text{III\_}}{\text{H}}_2\text{O}+{\text{NO}}_2^{-}\to {\text{Mb}}^{\text{III\_}}\text{nitrite}+{\text{H}}_2\text{O}$$ (5)

$${\text{Mb}}^{\text{IV}}=\text{O}+\text{NO}\to {\text{Mb}}^{\text{III\_}}\text{ONO}$$ (6)

Solution studies:

In this current work, we added excess nitrite to the ferric Mb^III–H₂O precursors to generate the Mb^III–nitrite derivatives. UV-vis spectral monitoring of the reaction (eq. 5) reveals changes in the spectra similar to those reported previously¹²¹ for the reaction of ferric wt sw Mb^III–H₂O with nitrite. Specifically, addition of sodium nitrite to a solution of ferric sw Mb^III–H₂O in phosphate buffer at pH 7.4 results in a red-shift of the Soret band from 408 nm to 409 nm, as well as a change in the band profile in the 450–650 nm region (Figures 4A and S1). The absorbance changes were plotted versus the concentration of nitrite to fit to the hyperbolic equation for a 1:1 binding model, with a K_D of 12.2 ± 2.2 mM.

Figure 4.

UV-vis spectral monitoring of the reactions of ferric wt sw Mb^III–H₂O. (A) changes after addition of nitrite. (B) changes after addition of dithionite followed by the addition of nitrite as well as the plot of ΔAbs(421–457) against time. (C) UV-vis spectra of the three forms of wt sw Mb (ferric, deoxy ferrous, and NO) with the enlarged Q region. Conditions: 0.1 M phosphate buffer, pH 7.4, [protein] = 1.8–7.4 μM, [nitrite] in (A) = 40 mM, [dithionite] = 5 mM, [nitrite] in (B) and (C) = 4 mM.

In contrast, no significant changes in the UV-vis spectra were observed when nitrite was added to the ferric H64A Mb^III–H₂O mutant even with concentrations up to 80 mM in [nitrite] (Figures S2A and S3), indicating very weak binding of nitrite to the heme. Spectral changes for the H64Q mutant are shown in Figures S4A and S5 with an observed red-shift of the Soret band from 408 nm to 412 nm; a K_D of 1.45 ± 0.30 mM was obtained. The related spectral changes for the V68A/I107Y mutant and resulting K_Ds are shown in Figures S6A and S7, and Table 2, respectively. The dissociation constants for nitrite binding to the heme centers of the wt and mutant Mb^IIIs thus increase in the order H64Q < wt < V68A/I107Y < H64A.

Table 2.

Dissociation constants for the binding of nitrite to the ferric sw wt and mutant Mb^III–H₂O proteins.

Protein	KD (mM)a
wt	12.2 ± 2.2
H64Q	1.45 ± 0.30
V68A/I107Y	19.4 ± 0.6
H64A	>> 20

^aIn 0.1 M phosphate, pH 7.4, at 25 °C.

Crystallographic studies:

Much controversy still surrounds the binding mode of the nitrite anion to Mb. As mentioned in the Introduction, heme protein–nitrite complexes generally display the N-binding mode of nitrite in their X-ray crystal structures.^18–31 Our published crystal structures for hh wt Mb^III–nitrite (and for human Hb^III) displaying the nitrite O-binding mode were thus unexpected.^{34, 36–38, 48} Most computational studies have also supported an N-binding ground state conformation in heme-nitrite complexes,^{19, 31, 39–46} although a recent thorough computational analysis supports the crystallographically determined O-binding mode in ferric Mb^III–ONO.⁴⁷ We thus sought to determine the X-ray crystal structures of the wt, H64A, H64Q, and V68A/I107Y Mb^III–nitrite complexes to examine if the distal pocket changes in these sw Mbs would affect their nitrite binding modes.

The four wt, H64A, H64Q, and V68A/I107 Mb^III–nitrite adducts were prepared by soaking nitrite into droplets containing crystals of the respective Mb^III–H₂O precursors for 20–30 mins. The crystal structures revealed the O-binding modes of the nitrite ligands in all four Mb^III–nitrite adducts as shown in Figure 5, with the nitrite ligands refined at 100% occupancy in each structure.

Figure 5.

The F_o–F_c omit electron density maps (green mesh, contoured at 3σ) and final models of the heme active sites of the ferric sw Mb^III–nitrite adducts. (A) wt Mb^III–nitrite at 1.85 Å resolution (PDB ID: 5UT7), (B) H64A Mb^III–nitrite at 1.85 Å resolution (PDB ID: 5UT9), (C) H64Q Mb^III–nitrite at 1.81 Å resolution (PDB ID: 5UTA), (D) V68A/I107Y Mb^III–nitrite at 1.57 Å resolution (PDB ID: 5UTD). The 2F_o–F_c maps are shown in Figure S9.

The O-binding mode in the sw wt Mb^III–nitrite complex (Figure 5A), with a H-bond between the nitrite O1 atom (i.e., Fe–ONO) and the distal His64 residue (O1…Nε(His64) = 2.8 Å), is analogous to that determined for the hh Mb^III–nitrite and human Hb^III–nitrite derivatives.^{34, 36–38, 48, 98, 122} The distance between the nitrite terminal O2 atom and Nε(His64) is 3.2 Å, with the next closest distances between the nitrite and distal pocket residues being that with the Val68 residue; at ~3.3 Å between the nitrite N and O1 atoms to Cγ2 of Val68, and at ~3.4 Å between the nitrite O2 atom and Cγ2 of Val68. Selected geometrical data for the heme–nitrite moieties for this and the other structurally characterized Mb–NOx derivatives from this work are collected in Table S2. Only minor structural changes in the heme pocket occur when nitrite replaces the bound H₂O ligand in wt Mb^III–H₂O to give the Mb^III–nitrite adduct (Figure 6A).

Figure 6.

Comparisons of the heme pockets of the sw ferric Mb^III–H₂O (in green), ferrous deoxyMb^II (in magenta) and ferric Mb^III–nitrite (in cyan) adducts. A) the wt adducts (rmsd 0.51 Å), (B) the H64A adducts (rmsd 0.13 Å), (C) the H64Q adducts (rmsd 0.14 Å), and (D) the V68A/I107Y adducts (rmsd 0.70 Å).

Interestingly, whereas an N-nitro binding mode was observed in the sw H64V Mb^III–nitrite adduct (lacking the distal His64 residue; Figure 2, right), the newly determined H64A sw Mb^III–nitrite structure (also lacking the distal His64 residue) unexpectedly reveals an O-binding mode of the nitrite anion to the heme Fe center (Figure 5B). Fixed water molecules in the vicinity of the distal pocket near the protein exterior provide a “water bridge” H-bonding interaction with the nitrite ligand with a (nitrite)O1…OH₂ distance of 2.9 Å. This “water bridge” is also present in the crystal structure of the H64A Mb^III–H₂O precursor. In this H64A Mb^III–nitrite complex, the related closest interaction between the nitrite ligand and distal pocket residues is with the Cζ of Phe43 ((nitrite)O2…Cζ = 3.5 Å); the Val68 sidechain is flipped from its original horizontal orientation in its H64A Mb^III–H₂O precursor to a vertical orientation in the nitrite adduct. Importantly, the apparent discrepancy between the observed N-nitro binding in the sw H64V Mb^III–nitrite adduct (Figure 2, right) and the sw H64A Mb^III–nitrite adduct reported here can be rationalized by the observation that the crystal structure of the precursor H64A sw Mb^III–H₂O contains an Fe-bound H₂O ligand;⁹⁷ thus, the O-bound nitrite appears to essentially simply replace another O-bound ligand (H₂O) as illustrated in Figure 6B. In contrast, however, the sw H64V Mb^III structure does not contain a fixed H₂O ligand to Fe,⁷⁹ implying there may be a lower affinity for an added O-bound ligand such as nitrite in the O-nitrito form.

Introducing a more flexible Gln distal residue in the H64Q mutant also generates an O-bound nitrite conformation as shown in Figure 5C. The distances between the O1 and O2 atoms of the bound nitrite and the N atom of Gln64 are 2.9 Å and 3.3 Å, respectively. Two conformations (horizontal and vertical) of the Val68 sidechain were modeled to fit the electron density in this nitrite adduct (Figures 5C and 6C). Such Val68 conformation flips have been observed previously in organometallic Mbs with direct Fe–aryl bonds.¹²³

Contrary to our initial expectations, the mutation of I107 to Tyr to provide a second H-bonding residue in the distal pocket (Figure 3, right) did not alter the O-binding mode of the nitrite ligand to an anticipated N-bound form (Figure 5D) in this V68A/I107Y Mb^III–nitrite adduct. As seen in Figure 6D, the His64 residue in this double-mutant swings inward (Nε movement of ~1.1 Å) with retention of the H-bond with the nitrite O1 atom as was the case in the wt Mb^III–ONO derivative; the O1 and O2 distances from Nε(His64) being 2.8 Å and 3.1 Å, respectively. The other closest contacts between nitrite and distal pocket residues are with the nearby Phe43 and Ala68 (at ~3.4 Å with the closest C atoms of these residues). The closest interaction between the nitrite ligand and the Tyr107 sidechain is between the nitrite O2 atom and the Oη of Tyr107 at a distance of 3.5 Å, a distance too long to form a substantial H-bond interaction.

Nitrite reduction to generate the nitrosyl Mb^II–NO derivatives

Nitrite reduction to NO mediated by deoxyMb^II has been reported for sw wt Mb (eqs. 3 and 4),⁵⁵ H64A mutant sw Mb,⁵⁵ and other sw and hh Mb proteins.^{48, 55, 66} Mechanisms of

$${\text{Mb}}^{\text{II}}+{\text{NO}}_2^{-}+{\text{H}}^{+}\to \text{NO}+{\text{Mb}}^{\text{III}}+{\text{OH}}^{-}$$ (3)

$${\text{Mb}}^{\text{II}}+\text{NO}\to {\text{Mb}}^{\text{II}}\text{NO}$$ (4)

such heme-based nitrite reduction have normally invoked the N-nitro bound form of nitrite as the starting point in the reaction pathway (Figures 1 and 7 (left)). Silaghi-Dumitrescu first proposed the possibility of a viable O-nitrito intermediate for heme-based nitrite reduction which would involve NO release and Fe–OH formation (Figure 7, right).⁴⁰ This concept has been extended to the possible involvement of a viable O-nitrito intermediate in Hb-based nitrite reduction.⁴³ In either pathway, the final end-product is the ferrous nitrosyl Mb^II–NO derivative (eqs. 3–4).

Figure 7.

Proposed pathways for heme-based nitrite reduction involving the N-nitro (left) and O-nitrito (right) isomers.

Solution studies:

Following established procedures, we performed the nitrite reduction experiments with the sw Mb^II H64Q and V68A/I107Y mutants. The ferric Mb^III–H₂O precursors were first reduced to their deoxyMb^IIs using dithionite followed by reaction with nitrite.

The reactions of nitrite with the H64Q and V68A/I107 deoxyMb^IIs in phosphate buffer at pH 7.4 and at 37 °C are illustrated in Figure S8, and yielded bimolecular reaction rates of 0.57 ± 0.06 M⁻¹s⁻¹ (H64Q) and 6.3 ± 0.4 M⁻¹s⁻¹ (V68A/I107). These may be compared with the reaction rates for wt sw Mb (5.6 ± 0.6 M⁻¹s⁻¹) and H64A sw Mb (1.8 ± 0.3 M⁻¹s⁻¹) reported previously.⁵⁵

The UV-vis spectra of the wt deoxyMb^II and Mb^II–NO derivatives are shown in Figures 4B-C; the mutant deoxyMb^IIs and Mb^II–NO spectra are shown in Figure S2 (H64A), Figure S4 (H64Q), and Figure S6 (V68A/I107Y). With these wt and mutant sw Mb^II–NOs in hand, we then sought to characterize them by X-ray crystallography to determine distal pocket effects on the heme–NO geometry.

Crystallographic studies:

Despite the biological relevance of Mb^II–NO complexes, there are only four reported X-ray crystal structures of Mb^II–NO derivatives.^{61,34, 62, 63} Brucker et al. first reported the crystal structure of wt sw Mb^II–NO at 1.7 Å resolution (∠FeNO = 112°) prepared from gaseous NO addition (after sodium fluoride addition) to crystalline Mb^III–H₂O;⁶¹ the preliminary structure of the L29F mutant Mb^II–NO derivative (∠FeNO = 130°) was only briefly mentioned. We reported the crystal structures of wt hh Mb^II–NO with varied but reproducible ∠FeNO angles of ~120° or ~144° dependent on the preparative conditions.^{34, 62} The crystal structure of blackfin tuna wt Mb^II–NO was also reported, with an ∠FeNO angle of 134°.⁶³ A few related ferrous Hb^II–NO structures are also known.^{64, 65} In any event, the availability of the four sw Mb^II–NO derivatives prepared in this work led us to determine their crystal structures to systematically examine distal pocket effects on the FeNO conformations in this series.

All four wt, H64A, H64Q, and V68A/I107Y Mb^II–NO derivatives were prepared from dithionite reduction of the in situ generated crystalline ferric Mb^III–nitrite precursors described above. Their crystal structures were determined, and the heme sites are shown in Figure 8. Modeling the FeNO group as nitrosyl (Fe–N–O) rather than isonitrosyl (Fe–O–N) stems from spectroscopic and theoretical (QM, QM/MM) results on Mb^II–NO^{124, 125} and model heme–NO systems^126–128 that show the isonitrosyl linkage isomers to be highly unstable even at low temperature and ~18–37 kcal/mol higher in energy.

Figure 8.

The F_o–F_c omit electron density map (green mesh, contoured at 3σ) and final model of the heme active site of the ferrous sw Mb^II–NO derivatives. (A) wt Mb^II–NO at 1.70 Å resolution (PDB ID: 5JNG), (B) H64A Mb^II–NO at 1.78 Å resolution (PDB ID: 5JNI), (C) H64Q Mb^II–NO at 1.78 Å resolution (PDB ID: 5JNJ), (D) V68A/I107Y Mb^II–NO at 1.79 Å resolution (PDB ID: 5JNK). The 2F_o–F_c maps are shown in Figure S10.

The sw wt Mb^II–NO heme site structure (Figure 8A) exhibits an ∠FeNO bond angle of 138°, which is consistent with the canonical ∠FeNO bond angle (138.5°) expected for synthetic six-coordinate ferrous nitrosyl porphyrins with N-base trans ligands,¹²⁹ and is also consistent with the previous quantum chemically predicted value (138.1°) for Mb^II–NO.¹³⁰

The ∠FeNO bond angle of 138° in sw wt Mb^II–NO prepared from the nitrite/dithionite method is larger than that observed for the complex prepared from the NO gas method (112°)⁶¹ but is closer to the 144° seen in ferrous hh Mb^II–NO prepared from the nitrite/dithionite method.³⁴ A H-bonding interaction between the nitrosyl N-atom and the distal His64 residue is present in this structure, with a (nitrosyl)N… Nε(His64) distance of 3.1 Å. The next closest distances between the NO ligand and distal pocket residues are with the Cγ2(Val68) atom; 3.2 Å distance from nitrosyl O atom, and 3.3 Å from the nitrosyl N-atom. Selected geometrical parameters for the wt and mutant sw Mb^II–NO derivatives described below are listed in Table S2.

The NO ligand in the H64A Mb^II–NO derivative is oriented towards the hydrophilic distal pocket exterior, in the general direction of the heme propionates (Figure 8B). This NO orientation towards the protein exterior is unprecedented in Mb^II–NO and Hb^II–NO protein structures, with no example reported in the literature or in the Protein Data Bank to date. In the absence of the H-bonding distal H64 residue, three fixed water molecules provide a H-bonding network from the nitrosyl O-atom to the protein exterior, with a (nitrosyl)O…OH₂ distance of 2.5 Å. The nitrosyl N and O atoms are 3.0 Å and 3.2 Å, respectively, from the Cγ2(Val68) atom.

The crystal structure of the H64Q Mb^II–NO derivative (Figure 8C) also reveals that the NO ligand is oriented towards the hydrophilic exterior, similar to that seen for the H64A derivative described above. Although the Gln64 residue provides a H-bonding interaction with the bound NO ligand, this H-bonding interaction is with the nitrosyl O-atom rather than the N-atom, which is quite unusual for Mb-like heme protein–NO derivatives. Interestingly, such heme nitrosyl ambidentate O-atom vs. N-atom H-bonding in heme models has been calculated to have an effect on Fe–N(O) and FeN–O vibrational spectral (normal vs. inverse) correlations.¹³¹

The structure of the V68A/I107Y Mb^II–NO derivative is similar to that of the wt Mb^II–NO in terms of the orientation of the NO ligand that points towards the hydrophobic protein interior (Figure 8D). Both the nitrosyl N and O-atoms are at a 2.9 Å distance from the Nε atom of His64. However, similarly to what was observed in the nitrite adduct, the Tyr107 sidechain is not in H-bonding contact with the NO ligand. The closest distance is between the nitrosyl O-atom and the Tyr sidechain O-atom (at 4.2 Å). Apart from the His64 residue, the closest distance between the NO ligand and distal pocket residues is 3.6 Å, between the Cβ atom of Ala68 and the nitrosyl O-atom.

It is thus interesting to note that for the four Mb^II–NO structures reported in this work, the two that contain the distal His64 residue display NO ligand orientations in the direction of the hydrophobic protein interior (i.e., away from the heme propionates). In contrast, the two mutants that do not contain the distal H64 residue (for any Mb and Hb structures) display unprecedented NO ligand orientations towards the hydrophilic exterior. These observations are summarized in the heme site structural overlays shown in Figure S11.

We found that this orientational preference for the NO ligand in the wt and mutant Mb^II–NOs in this work was reproducible over the several (>10) independent compound preparations, using two different orders of addition of dithionite and nitrite, and X-ray diffraction data collection experiments. Silvernail et al.¹³² have determined, using computational methods for a series of six-coordinate model (por)Fe(NO)(imidazole) compounds, that there is no significant difference in potential energy between the two observed (opposite) NO orientations in the simple non-protein models. With an expanded set of protein Mb^II–NO derivatives with varying NO ligand orientations available to us, we proceeded to probe these orientational preferences with both QM and QM/MM methods.

Computational studies: Probing the experimentally observed differential FeNO orientations:

In order to help understand why the NO adopts different orientations in the different wt and mutant Mb^II–NO derivatives, we probed, using QM and QM/MM calculations, both the original NO orientations as observed in the experimentally derived X-ray crystal structures and the alternative NO orientation models. The active site models of the X-ray crystal structures and optimized geometries are shown in Figure 9. The relative energies of the systems containing the hypothetical alternative NO orientations (i.e., compared with the experimentally observed X-ray orientations) are listed in Table 3. For the wt Mb^II–NO, the QM optimized structure with the X-ray NO orientation (wt_b in Figure 9) is similar to that in the original X-ray structure (wt_a), with the H-bonding between NO and distal His64 maintained in the optimized structure. The QM optimized structure with the alternative NO orientation (wt_c), however, reveals significant differences. In particular, the H-bond to NO is lost (in wt_c) and there is also ample displacement regarding the relative positions of the distal pocket residues His64 and Val68.

Figure 9.

(a) Active site models from X-ray structures. (b) QM-X: QM optimized structures using X-ray orientations of NO. (c) QM-A: QM optimized structures of alternate NO orientations. (d) QM/MM-X: QM/MM optimized structures using X-ray orientations of NO. (e) QM/MM-A: QM/MM optimized structures of alternate NO orientations.

Table 3.

Relative energies (in kcal/mol) and Fe–NO bond lengths (in Å) of proteins with the alternative NO orientations compared to the X-ray orientations in the wild-type (wt) and mutant Mb^II–NO derivatives.

Method	Protein	Figure 9 labela	ΔE	ΔEZPE	ΔH	ΔG	ΔRFe-NO
QM	MbII–NO	wt_c	0.47	0.26	–0.75	1.82	0.026
	H64A MbII–NO	H64A_c	5.17	3.57	3.39	3.26	0.009
	H64Q MbII–NO	H64Q_c	–0.19	–0.41	–0.93	0.82	0.009
	V68A/I107Y MbII–NO	AY_c	0.60	0.48	1.09	–2.01	–0.001
QM/MM	MbII–NO	wt_e	/	/	/	/	/
	H64A MbII–NO	H64A_e	22.63	19.06	21.34	11.60	0.006
	H64Q MbII–NO	H64Q_e	4.09	3.35	3.85	3.86	0.001
	V68A/I107Y MbII–NO	AY_e	0.09	6.56	6.41	5.19	0.004

^aStructure labels as indicated in Figure 9.

These features result in a Gibbs free energy penalty of 1.82 kcal/mol for this alternative NO orientation (Table 3). As this alternative orientation triggers big changes in the positions of the His64 and Val68 residues in these truncated active site models, which may not be possible to be accommodated in the actual protein due to the protein environment, attempts at the QM/MM geometry optimization of the whole protein with this alternative orientation, indeed, led to the same observed X-ray orientation (i.e., wt_e = wt_d).

For the H64A mutant (H64A_a in Figure 9), the QM optimized X-ray NO orientation (H64A_b) structure with the NO ligand pointing towards the protein exterior is close to the original X-ray structure (H64A_a), while the alternative NO orientation optimized structure (H64A_c; with NO pointing towards the hydrophobic interior) reveals large differences. In particular, the original H-bond between NO and the active site water molecule is again lost (H64A_c); similar to the loss of H-bonding with the distal His64 residue in wt Mb^II–NO (wt_c). For this system, the alternative orientation (H64A_c) not only costs 3.26 kcal/mol in Gibbs free energy, but also has a significant cost around 3–5 kcal/mol in other energy terms. Consistent with this relatively larger energy difference compared with the wt Mb^II–NO case, the positions of three more active site residues are significantly affected, including Phe43, Thr67, and Ile107, besides Ala64 and Val68 corresponding to His64 and Val68 in wt Mb^II–NO. This extensive effect on the active site residues eventually affects the whole protein structure more significantly, with the QM/MM optimized alternative NO orientation (H64A_e) being less stable than the X-ray orientation (H64A_d) by 11.60 kcal/mol in Gibbs free energy. The large difference in the QM versus the QM/MM results emphasize the importance of considering the whole protein even when focusing on only heme site ligand conformations.

In contrast with above obvious favorable energies in the QM calculations for the X-ray NO orientations for wt and H64A Mb^II–NO compounds, the H64Q Mb^II–NO derivative has a Gln64 residue that is relatively more flexible than His64 in forming H-bonds. Indeed, NO can form a H-bond with Gln64 in both the X-ray NO orientation (H64Q_b; H-bond with the nitrosyl O atom) and the alternative orientation (H64Q_c; H-bond with the nitrosyl N atom). However, the H-bond in the X-ray orientation (H64Q_b; involving the nitrosyl O atom) is still slightly more favorable, as indicated by a shorter H-bond distance of 2.250 Å vs. 2.476 Å in the alternative orientation. Accordingly, the X-ray orientation active site structure is more stable by 0.82 kcal/mol in Gibbs free energy. The corresponding QM/MM optimized structure with the X-ray orientation (H64Q_d; H-bond with the nitrosyl O atom) is also more favorable than the alternative orientation (H64Q_e; H-bond with the nitrosyl N atom) by a more significant energy difference of ΔG of 3.86 kcal/mol, but is still the smallest penalty for the alternative NO orientation among all QM/MM calculations here (Table 3).

For the double mutant V68A/I107Y Mb^II–NO, the QM optimized alternative orientation (AY_c; H-bond between His64 and the nitrosyl O atom) is surprisingly more stable than the X-ray orientation (AY_b; H-bond with the nitrosyl N atom) by ΔG of 2.01 kcal/mol. The H-bond distances for NO with His64 in both structures are similar; 2.457 Å (AY_b; X-ray orientation) vs. 2.499 Å (AY_c; calculated alternative orientation). However, a closer look at the optimized structures (AY_b vs. AY_c) in the truncated models used in the QM calculations reveals unexpected large differences in the positions of Phe43 and Tyr107 in the optimized X-ray orientation structure (AY_b) when compared with the experimentally determined X-ray structure (AY_a). Nevertheless, once the whole protein was used, the QM/MM calculations show a clear energy preference for the X-ray orientation (AY_d; H-bond with the nitrosyl N atom) by ΔG of 5.19 kcal/mol (Table 3). In addition, these active site residues’ positions are now closer to the X-ray structure (AY_a) than those from the QM truncated model studies (AY_b). The H-bond distance of NO with His64 of 2.146 Å in the X-ray orientation (AY_d) is also more favorable than that (at 2.335 Å) in the alternative orientation (AY_e). Results for this protein system suggests that one should be careful about results in truncated model studies and sometimes the large protein models are necessary to obtain more reliable data.

Overall, above investigations show that NO prefers the orientations as observed in the X-ray structures with more stable energies. Interestingly, as shown in Table 3, in the X-ray orientation, the Fe–NO bond length is relatively shorter (ΔR_Fe-NO) than that in the alternative orientation from both QM and QM/MM calculations with correct energy orders. This again suggests a more stable NO binding. In addition, these results indicate that when there is a non-His64 distal residue such as Ala64 or Gln64, NO adopts an orientation toward this residue, while if there is a His64 residue present, NO prefers an orientation point away from this residue in the direction of the protein interior, to achieve more stable binding overall in the protein environment.

CONCLUSIONS

In this work, we have clarified the preferred O-nitrito binding of nitrite to the heme center in wt and mutant Mb^IIIs differing in the number of potential H-bonds (i.e., 0, 1, or 2) to the bound ligand. The dissociation constants increase in the order H64Q < wt < V68A/I107Y < H64A. This well-defined O-nitrito binding to Mb^IIIs (and Hb^III) remains unique in heme-nitrite structural biology and presents opportunities for further nitrite structural biology using other heme proteins with different active site architectures. Replacement of the bound H₂O ligand in the Mb^III–H₂O precursors with the O-bound nitrite does not result in substantial changes in the overall Mb protein fold. Minor changes in side-chain orientations in the active site were, however, observed, such as the orientations of the Val68 residue in the H64A and H64Q mutants (i.e., without the distal His64 residue) and the inward movement of the His64 residue in the V68A/I107 double mutant derivative to maintain its H-bond with the bound nitrite.

Nitrite reduction to NO is enabled by the ferrous Mb^IIs to generate the Mb^II–NO derivatives. Importantly, we report the first systematic structural investigation of the FeNO geometries in wt and mutant Mbs that led to unexpected but reproducible FeNO conformational preferences in these proteins. We found that when the natural His64 residue was present (wt and V68A/I107Y), the NO vector was pointed towards the hydrophobic protein interior as previously noted for other wt Mb^II–NO and Hb^II–NO compounds. However, in the absence of this His64 residue (H64A and H64Q), the NO vector was oriented towards the hydrophilic protein exterior. Detailed QM and hybrid QM/MM theoretical investigations were employed to rationalize these varied FeNO orientations observed experimentally by X-ray crystallography. In addition, these calculations underpin the caution that in probing such FeNO orientations in heme proteins, sometimes the whole protein needs to be taken into account when examining such heme-ligand conformational preferences in protein derivatives.