New Light on NO Bonding in Fe(III) Heme Proteins from Resonance Raman Spectroscopy and DFT Modeling

Abstract

Visible and ultraviolet resonance Raman (RR) spectra are reported for Fe^III(NO) adducts of myoglobin variants with altered polarity in the distal heme pockets. The stretching frequencies of the Fe^III–NO and N–O bonds, ν_FeN and ν_NO, are negatively correlated, consistent with backbonding. However, the correlation shifts to lower ν_NO for variants lacking a distal histidine. DFT modeling reproduces the shifted correlations, and shows the shift to be associated with the loss of a lone-pair donor interaction from the distal histidine that selectively strengthens the N–O bond. However, when the model contains strongly electron-withdrawing substituents at the heme β-positions, ν_FeN and ν_NO become positively correlated. This effect results from Fe^III–N–O bending, which is induced by lone pair donation to the N_NO atom. Other mechanisms for bending are discussed, which likewise lead to a positive ν_FeN/ν_NO correlation, including thiolate ligation in heme proteins and electron-donating meso-substituents in heme models. The ν_FeN/ν_NO data for the Fe(III) complexes are reporters of heme pocket polarity and the accessibility of lone pair, Lewis base donors. Implications for biologically important processes, including NO binding, reductive nitrosylation and NO reduction, are discussed.

Introduction

Heme protein sensors^1–3 of the gaseous ligands CO, NO and O₂ play key regulatory roles in biology. Other heme proteins catalyze oxidative or reductive chemistry of the same ligands. The nature of the heme-ligand adducts and their interactions with the surrounding protein are of continuing interest.

Among these ligands, NO is unique in being able to bind heme in both Fe(II) and Fe(III) oxidation states. The odd electron on NO gives it flexible electronic properties. Fe^III(NO) adducts are often transient, being easily reduced to the more stable Fe^II(NO) adducts. However, Fe^III(NO) is stabilized in some heme proteins. These include the nitrophorins,⁴ used by blood-sucking insects to release NO as a vasodilator into the victim’s blood stream. Maintaining the Fe(III) oxidation state assures reversible binding and release of NO. Fe^III(NO) is also involved in the catalytic cycle of NO synthases,⁵ and of NO redox enzymes, including nitrite reductase⁶ and NO reductase.⁷

The dominant mechanism for XO binding to heme (X = C, N, O) is backbonding, because the ligands have low-lying π* orbitals, well matched to the filled d_π orbitals on Fe(II). These orbitals are contracted on Fe(III), which is unable to bind CO or O₂. However, the odd electron on NO can readily transfer to Fe(III), producing the Fe^II(NO⁺) resonance structure, isoelectronic with Fe^II(CO). DFT computations on model heme adduct show that Fe^II(NO⁺) is indeed the appropriate description of the Fe^III(NO) ground state.⁸ However, Lehnert and coworkers report computation of a low-energy Fe^III(NO) radical state, in which the Fe^III–NO bond is slightly elongated and the odd electrons on Fe(III) and NO are antiferromagnetically coupled.⁸ This state is likely involved in the mechanism of Fe^III–NO dissociation. Walker discussed this state previously and pointed out that it should be stabilized by porphyrin ruffling, a structural feature of the nitrophorins.⁴

Backbonding produces anticorrelation of the Fe–XO and X–O bond strengths, in response to polarization effects. Enhancement of backbonding strengthens Fe–XO while weakening X–O. Negative correlations between Fe–XO and X–O stretching frequencies, gathered from Raman and/or IR spectra, have been observed for Fe(II) adducts of all three XO ligands.⁹ However, Fe^III(NO) adducts are exceptional. When the available literature data are plotted (Figure 1), the dominant ν_FeN/ν_NO correlation seems to be positive; the Fe^III–NO and N–O bond strengths increase or decrease in concert for oxidized iron. Rodgers and coworkers^10,11 have investigated this phenomenon computationally, and have found positive ν_FeN/ν_NO correlations when electron donating and withdrawing heme meso-substituents were used to polarize the Fe^III(NO) adduct, a technique used previously,^12,13 both experimentally and computationally to explore the negative correlations found for Fe^II(XO) adducts. They attributed the contrary behavior of Fe^III(NO) to a unique occupied molecular orbital that is antibonding with respect to both the Fe^III–NO and N–O bonds.^10,11

Figure 1.

ν_FeN/ν_NO data for six-coordinate Fe^III(NO) heme proteins and model complexes (see Table 1). “” indicates histidine-ligated heme-protein adducts, while “” indicates cysteine-ligated heme-protein adducts; “” and “” represent Fe^III(NO) porphyrin model complexes. The blue line is a linear correlation for thiolate-ligated adducts, with r = 0.84, after excluding the data for model complexes (SR and SR-HB).

However, the available experimental data are sparse (Table 1), owing to the limited stability of Fe^III(NO) adducts. The data describing the positive correlation in Figure 1 are all from thiolate-ligated heme proteins and models. The available data for non-thiolate adducts appear as a scattered cluster in the upper right corner of the figure. In order to examine this cluster more closely, we have obtained a new data set from resonance Raman (RR) spectra of a series of myoglobin (Mb) variants, used previously to explore backbonding in Fe^II(CO)¹⁴ and Fe^II(NO)¹⁵ complexes (Figure 2). These data reveal negative ν_FeN/ν_NO correlations for Fe^III(NO) adducts, fully consistent with backbonding. They also reveal a specific N–O strengthening effect that is associated with the distal histidine of Mb. DFT modeling confirmed the latter effect, showing it to result from a lone pair interaction of the imidazole with the positively charged Fe^II(NO⁺), an effect that was originally postulated by Miller et al.¹⁶ Further modeling revealed that a positive ν_FeN/ν_NO correlation stems from Fe^III–N–O bending, which can be induced by several mechanisms, including thiolate ligation. Implications of these results for heme protein function are discussed.

Table 1.

Fe^III–NO and N–O Stretching Frequencies (cm⁻¹) for Six-Coordinate Fe^III(NO) Complexes of Heme Proteins and Porphyrin Models.

Protein/porphyrin	ν(Fe–NO)	ν(N–O)	Refs
(His)FeIII(NO) complexes
NORa	594	1904	7
HRPb	604	1903	17, 18
hHO-1c	596	1918	19
NP1d (pH 8.0)	591	1917	20, 21
NP1 (pH 5.6)	591	1904	20, 21
NP4e (pH 7.5)	593	1921	22, 23
NP4 (pH 5.5)	590	1904	22, 23
HbNf	591	1914	24
HbN-Y33Fg	592	1908	24
HbAh	594	1925	17
Mbi	594	1921	t.w.p
OEP(Pyr)j	602	1917	25, 26
(Cys)FeIII(NO) complexes
P450camk (substrate free)	528	1806	27, 28
P450cam+camphor	522	1806	27, 28
P450cam+norcamphor	524	1818	27, 28
P450cam+adamantanone	520	1818	27
CPOl	538	1868	27, 28
P450norm	529	1851	28, 29
P450nor (proto)	530	1853	30
P450nor (meso)	532	1852	30
P450nor (deutero)	533	1851	30
SRn	510	1828	31
SR-HBo	515	1837	31

^aBacterial nitric oxide reductase from Paracoccus denitrificans.

^bHorseradish peroxidase.

^cHuman heme oxygenase-1.

^dNitrophorin 1 from Rhodnius prolixus.

^eNitrophorin 4 from Rhodnius prolixus.

^fTruncated hemoglobin from Mycobacterium tuberculosis.

^gTyrosine 33 to phenylalanine mutant of HbN.

^hHuman hemoglobin.

ⁱMyoglobin.

^jFe(III)-octaethylporphyrin with pyridine as proximal ligand.

^kCytochrome P450cam.

^lChloroperoxidase.

^mCytochrome P450 nitric oxide reductase from Fusarium oxysporum.

ⁿFe(III) porphyrin-alkanethiolate complex.

^oHydrogen-bonded SR.

^pThis work.

Figure 2.

Structure of sperm whale Mb^II(CO) (pdb ID: 1A6G), showing the mutated distal residues in the variants studied.

Materials and Methods

Mb Samples

Preparation of the Mb variants has been described elsewhere.^32,33 Mb samples were converted to the Fe(III) (metMb) form by adding ca. 10 fold excess of sodium ferricyanide (Aldrich), and removing excess ferricyanide with a Sephadex G-25 PD-10 column (GE Healthcare), pre-equilibrated with 100 mM sodium phosphate, pH 7.0 buffer. Fe^III(NO) adducts were then prepared either by passing ¹⁴NO gas (Matheson Gas Products) anaerobically through a 200 μL, 0.1–0.2 mM sample in a sealed NMR tube, or by injecting the required amount of chemically generated ¹⁵NO via an air-tight syringe.

Resonance Raman Measurements

RR spectra were collected for the ν_FeN region using 413.1-nm excitation from a Kr⁺ ion laser (Spectra Physics, 2080-RS), and for the ν_NO region using 244-nm excitation from an intracavity doubled-Ar⁺ ion laser (Innova 300 FReD, Coherent Radiation, Palo Alto, CA). Photodissociation and reduction of the NO adducts were minimized by using low laser power (~1 mW at the sample) and by spinning the sample. For the visible RR experiments, the scattered light was collected and focused onto a triple spectrograph (Spex 1877) equipped with a CCD detector (Roper Scientific, Model 7375-0001) operating at −110 °C. Spectra were calibrated with dimethylsulfoxide-d₆. For the ultraviolet RR (UVRR) experiments, a single spectrograph (Spex 1269) equipped with a UV-enhanced CCD detector (Princeton Instruments, Model LN/CCD-1340/400) was used to collect the scattered light. UVRR spectra were calibrated using acetone and dimethylsulfoxide-d₆.

DFT Calculations

Density functional theory (DFT; B3LYP) calculations on the model complexes were performed using the Gaussian 03 program.³⁴ The standard 6-31G* basis set was used for all the atoms except Fe, for which Ahlrichs’ valence triple-ζ (VTZ) basis set was chosen.³⁵ Geometry optimization and frequency calculations were performed using tight convergence criteria and an ultrafine integration grid. C_s symmetry constraints were used for the Fe^III(NO) porphyrin models without distal imidazole, where proximal imidazole was kept in the plane bisecting the Fe–N(pyrrole) bonds. For the models with distal imidazole, structures were optimized under C_s or no symmetry (C₁) constraints as specified for each case. Vibrational frequency values were taken directly from the Gaussian program without scaling. In all calculations, a singlet ground state for the Fe^III(NO) complexes was assumed.

Results

Resonance Raman Spectra of Fe^III(NO) Mb Variants

Although a number of RR spectra for Fe^III(NO) heme adducts have been reported (Table 1), the data are sparse. Facile autoreduction to the more stable Fe^II(NO) adducts is undoubtedly responsible for the relative paucity of literature reports. Although we hoped to obtain systematic data on model complexes, as we had for Fe^II(CO)¹² and Fe^II(NO)^13,36 adducts, our efforts have been frustrated by autoreduction and high background scattering.

We have, however, obtained high quality visible and ultraviolet RR spectra of carefully prepared Fe^III(NO) adducts of Mb using low laser intensities. The samples included Mb variants with distal residue substitutions (Figure 2), designed to vary the electric field distribution around the bound NO.^14,37 These variations have been shown to produce systematic vibrational shifts in Fe^II(CO)³⁸ and Fe^II(NO)¹⁵ adducts of Mb, and we sought similar effects for the isoelectronic Fe^III(NO).

RR spectra are illustrated in Figure 3 for wild-type (WT) Mb. The spectrum obtained with 413.1-nm excitation, in resonance with the heme Soret absorption, reveals the Fe^III–NO stretching and bending vibrations at 594 and 571 cm⁻¹, respectively, assigned¹⁷ by the shifts induced upon substituting ¹⁵NO for ¹⁴NO. To detect the N–O stretching vibration, we utilized excitation at 244 nm, where enhancement of this mode occurs, probably via resonance with an Fe(III)←NO ligand-to-metal charge-transfer (LMCT) transition.¹⁸

Figure 3.

RR spectra for Fe^III(NO) adducts of wild-type Mb containing ¹⁴NO or ¹⁵NO, and the difference spectra (¹⁴NO – ¹⁵NO).

Spectra of similar quality were obtained for the remaining Mb variants; some of their ¹⁴NO – ¹⁵NO difference spectra are displayed in Figures 4 and 5, which reveal a novel and unexpected pattern. The ν_FeN bands all fall in a narrow range, 590–600 cm⁻¹, but the ν_NO bands occupy two distinct regions, one at ~1920 cm⁻¹, and one at ~1900 cm⁻¹. The common thread is that the distal histidine residue, His64, is present in the ~1920 cm⁻¹ variants (Figure 4), but is replaced by hydrophobic residues (Ile or Leu) in the ~1900 cm⁻¹ variants (Figure 5). However, two of the His64-containing variants, F46L and F46V, show both ν_NO bands. In these variants the bulky Phe46 residue, which buttresses the His64 side chain (Figure 2), is replaced by smaller residues. Other studies have shown that this substitution increases the mobility of His64, whose side chain can occupy alternative positions, an inward conformer neighboring the heme NO ligand or an outward conformer away from it.^14,39,40 Thus, the two ν_NO bands in these mutants reflect populations in which His64 does or does not interact with the bound NO.

Figure 4.

RR difference spectra (¹⁴NO – ¹⁵NO) for Fe^III(NO) adducts of the indicated Mb variants, which retain the distal His64.

Figure 5.

RR difference spectra (¹⁴NO – ¹⁵NO) for Fe^III(NO) adducts of the indicated Mb variants, in which distal His64 is replaced by hydrophobic residues leucine or isoleucine.

These data are listed in Table 2, along with data from additional variants (spectra in Figure S1), in which His64 is replaced by the polar residue Gln, or in which Arg45 is replaced with smaller or oppositely charged amino acids. Again the ~1900- and 1920-cm⁻¹ bands show up simultaneously. The R45A and R45E substitutions eliminate an anchoring salt-bridge between Arg45 and a heme propionate, and increase the mobility of His64,^14,41,42 thus accounting for the two populations. In the R45K variant, this salt-bridge is preserved by the positively charged Lys45, and only the WT-like population is seen. For the H64Q variants, we infer that there are again two populations, this time with Gln64 interacting or not with the bound NO, in the same manner as His64. The Gln64 side chain can donate a lone pair from the carbonyl O^ε atom, but is expected to be more mobile than that of His64.

Table 2.

Fe^III–NO and N–O Stretching Frequencies (cm⁻¹) of Six-Coordinate Fe^III(NO) Myoglobin Mutants.

Mb mutant	ν(Fe–NO)	ν(N–O) (%)a
WT	594	1921
V68F	590	1923
F46L	595	1920 (61), 1901 (39)
F46V	596	1920 (48), 1902 (52)
L29F	598	1918
H64I	597	1900
H64L	596	1898
H64L/V68F	595	1901
H64L/L29F	599	1899
H64L/V68N	598	1911
H64Q	597	1920 (44), 1903 (56)
H64Q/V68F	594	1921 (75), 1903 (25)
H64Q/L29F	598	1919 (36), 1900 (64)
H64Q/L29F/V68F	596	1919 (82), 1901 (18)
R45A	594	1923 (81), 1905 (19)
R45E	594	1923 (75), 1905 (25)
R45K	594	1922

^aFor Mb variants with two ν_NO frequencies, the percentage area for each band is shown in parentheses.

When ν_FeN is plotted against ν_NO (Figure 6), it is apparent that there are two parallel correlations (solid lines), one in which His64 or Gln64 interacts with the NO, and one in which this interaction is absent. Both correlations are negative, as expected for backbonding, and the values of the slopes, −1.1 and −1.2, respectively, are somewhat larger than those observed when ν_FeC is plotted against ν_CO for Fe^II(CO) adducts, −0.73.⁴³ Side chain substitutions that increase the positive polarity near the bound NO, e.g., via the Phe multipole in L29F,⁴⁴ shift the point higher on either line.

Figure 6.

ν_FeN/ν_NO data for Fe^III(NO) adducts of the indicated Mb variants, without () and with () lone-pair donation to the bound NO from a distal His64 or Gln64. Asterisks indicate minor components when two ν_NO populations are detected. The red and blue lines are linear correlations with r = 0.88 and 0.84, respectively.

DFT Modeling

Density functional theory was used to analyze trends in structure and vibrational frequencies for Fe^III(NO) heme adducts. The basic model was (ImH)Fe^III(NO)P, where ImH is imidazole and P is porphine. The level of theory was the same (see Material and Methods section) as was used successfully for Fe^II(CO) adducts.³⁶ The calculated structure (see bond parameters for FeP in Table 3) agrees well with those reported by others,^8,10,45 using different levels of theory, and also with available model complex crystallographic data.^46,47 No Mb^III(NO) crystal structure is available, but EXAFS analysis⁴⁸ has indicated a linear Fe^III–N–O unit, and bond distances consistent with the models.

Table 3.

Selected Bond Lengths (Å), Bond Angles (deg), and Vibrational Stretching Frequencies (cm⁻¹) Calculated for Six-Coordinate (ImH)Fe^III(NO), (ImH)Fe^III(NO)···(ImH) and (ImH)Fe^III(NO)···(H₂O) Adducts.^a

Complex	Fe–NO	N–O	∠FeNO	Fe–NImp	NNO···NImd	ONO···NImd	ν(Fe–NO)	ν(N–O)
(ImH)–FeIII–NO, Cs symmetry
FeP(NH2)8	1.630	1.138	178.4	2.027	–	–	666	2057
FeP(CH3)8	1.632	1.137	179.7	2.025	–	–	666	2063
FeP	1.636	1.136	179.8	2.020	–	–	660	2070
FePF8	1.642	1.134	179.8	2.019	–	–	649	2077
FePCl8	1.643	1.133	179.8	2.014	–	–	647	2079
FeP(NO2)8	1.656	1.130	179.9	2.005	–	–	629	2092
(ImH)–FeIII–NO←:N(ImH), Cs symmetry
FeP(NH2)8	1.642	1.132	175.4	2.026	3.454	2.855	650	2084
FeP(CH3)8	1.645	1.131	175.3	2.023	3.426	2.832	647	2088
FeP	1.650	1.130	175.8	2.020	3.401	2.777	639	2096
FePF8	1.657	1.128	177.3	2.019	3.353	2.706	626	2103
FePCl8	1.660	1.127	177.2	2.014	3.346	2.685	623	2104
FeP(NO2)8	1.683	1.126	178.5	2.005	3.264	2.535	575	2101
(ImH)–FeIII–NO←:N(ImH), C1 symmetry
FeP(NH2)8	1.638	1.133	176.1	2.028	3.351	2.966	656	2080
FeP(CH3)8	1.642	1.132	174.5	2.025	3.238	2.929	650	2085
FeP	1.648	1.131	172.2	2.021	3.118	2.920	641	2089
FePF4	1.654	1.130	169.4	2.019	3.010	2.921	627	2088
FePF8	1.662	1.130	166.8	2.016	2.922	2.917	609	2083
FePCl8	1.672	1.130	163.4	2.010	2.794	2.897	587	2072
FeP(NO2)8	1.713	1.131	156.0	2.003	2.529	2.797	519	2041
(ImH)–FeIII–NO←(:OH2), C1 symmetry
FeP	1.647	1.131	173.4	2.020	2.949	2.807	644	2092
(ImH)–FeIII–NO←(:OH2)2, C1 symmetry
FeP	1.655	1.126	174.6	2.022	2.97/3.12	2.74/2.78	635	2117
(ImH)–FeIIIPH–NO, constrained ∠FeIIINO, Cs symmetry
FeP	1.755	1.146	140°	1.981	–	–	440	1938
FeP	1.701	1.141	150°	1.989	–	–	505	1993
FeP	1.664	1.138	160°	2.002	–	–	594	2036
FeP	1.643	1.136	170°	2.014	–	–	645	2062
FeP	1.641	1.136	172°	2.017	–	–	649	2065
FeP	1.639	1.136	174°	2.018	–	–	657	2067

^aN_Im^p and N_Im^d are the nitrogen atoms of proximal and distal imidazoles, respectively (see Figure 9).

The computed ν_FeN and ν_NO frequencies, 660 and 2070 cm⁻¹, are overestimated, as is usually the case with DFT, for which empirical scaling factors are often applied.^49–51 The required scaling factor for ν_NO, ~0.92, is indeed the same as determined by Pulay and coworkers⁵⁰ and consistent with the range generally encountered for molecules of the first row elements.⁵¹ Scaling factors for metal–ligand bonds have not been collected. Since we are interested in frequency trends, and not in absolute values, we have not applied scaling factors.

(a) Inductive Effects of Peripheral β-Substituents

To examine the effects of backbonding, we resorted to the same method as employed previously for Fe^II(CO)^12,36 and Fe^II(NO)³⁶ adducts, attaching electron donating or withdrawing substituents to the porphine periphery. Electron donors are expected to increase backdonation to the axial ligand, via their inductive effect, while electron-withdrawing substituents should decrease backdonation. This computational device has been applied to (ImH)Fe^III(NO)P by Linder and Rodgers,¹⁰ who found a positive correlation of ν_FeN and ν_NO, contrary to the negative correlation expected for backbonding. However, the positive correlation was based on a set of substituents attached to the meso-carbon atoms of porphine (Figure 7). Their calculations for substituents attached to the β-carbon atoms did not follow this correlation, and indeed the β-substituent points appeared to be negatively correlated.

Figure 7.

Fe porphine with eight substituents on the β-carbon atoms.

For Fe^II(CO) adducts the position of substitution does not materially affect the slope of the ν_FeC/ν_CO correlation,³⁶ but for Fe^III(NO) there is a specific electronic effect of meso-substituents, because of an altered orbital energy structure, as discussed below. Consequently, we limited the position of substitution to the β-carbon atoms. Table 3 lists structures and frequencies for (ImH)Fe^III(NO)P with eight β-substituents, –NH₂, –CH₃, –H, –F, –Cl, and –NO₂, arranged from the most donating to the most withdrawing. The resulting ν_FeN/ν_NO correlation (Figure 8) gives a straight line with a slope of −1.1, in excellent agreement with the experimental correlations found for the Mb^III(NO) adducts. Thus, the experimental data behave according to the backbonding model.

Figure 8.

ν_FeN/ν_NO correlation plots computed for the six-coordinate (ImH)Fe^III(NO)PX₈ complexes, without (blue points) and with a distal ImH, held in C_s symmetry (red triangles). “○” indicates (ImH)Fe^III(NO)P with a distal water molecule. Linear correlations are shown for models without distal ImH (blue line, r = 0.98), and with distal ImH in C_s symmetry (red line, r = 0.98), but excluding the –NO₂ data point.

(b) Distal Imidazole Interaction

Next we modeled the effect of a distal histidine interaction, by placing an imidazole ring near the bound NO ligand. In the first set of calculations, the distal ImH was restricted to the same plane as the proximal ImH, which bisected the Fe–N(pyrrole) bonds (C_s overall symmetry, Figure 9a).

Figure 9.

DFT-optimized molecular structures and atom labeling for (ImH)Fe^III(NO)P with a distal ImH held in C_s symmetry (a), or allowed to relax to C₁ (b).

The distal ImH was initially placed with its N–H proton pointing at the NO, simulating H-bonding. This orientation is stable for Fe^II(CO)^37,52–54 and Fe^II(NO)^55,56 adducts, but for (ImH)Fe^III(NO)P no stable structure could be found with this orientation. However, a stable structure was found when the imidazole N: lone pair was oriented toward the NO (Figure 9a). The imidazole ring was canted to one side of the NO, but the (ImH)N: atom was much closer to the O_NO than the N_NO atom, 2.78 vs. 3.40 Å (Table 3).

This stable structure supports the suggestion advanced some time ago by Miller et al.¹⁶ that the distal His64 in Mb engages in a lone pair donor interaction with Fe^III(NO). Their suggestion was based on the observation, confirmed in the present work, that the ν_NO (measured via FTIR spectroscopy at 4 °C to prevent autoreduction; ν_FeN was not reported¹⁶) shifted down when His64 was replaced by a hydrophobic residue, whereas an upshift would have been the direction expected on the basis of backbonding, if His64 exerts positive polarity, as it does for Fe^II(CO).^9,14

Next we computed structures and frequencies for β-substituted porphines, with the distal ImH in the same C_s-symmetry orientation (Figure 9a). The imidazole N: moves farther from the O_NO atom as the substituents become more electron-donating, and the Fe^III–N–O unit bends slightly, in the direction away from the distal imidazole (Table 3). The N–O distance increases while the Fe–NO distance decreases, as expected from increased backbonding.

The computed frequencies now fall on a new ν_FeN/ν_NO correlation (slope = −1.3), nearly parallel to the model without a distal ImH, and displaced ~20 cm⁻¹ to higher ν_NO values (Figure 8). Thus, the pattern observed for the Mb^III(NO) variants (Figure 6) is reproduced remarkably well. For a given porphyrin, the effect of the distal imidazole is to increase ν_NO and decrease ν_FeN (Table 3). Qualitatively, this effect is consistent with decreased backbonding, but the ν_NO increase is more pronounced, producing the observed shift in the correlation. (We note that agreement with experiment is inexact in that the Mb data indicate no effect of the distal imidazole on ν_FeN. For variants in which the distal His is mobile, two values of ν_NO are observed, but only one ν_FeN.)

What accounts for the shifted correlation? Figure 10 compares selected occupied molecular orbitals, computed for the distal imidazole model in C_s symmetry. The two orbitals on the left, HOMO-3 and -4, are the nearly degenerate pair that describes Fe^II–NO⁺ backbonding, and are similar to the backbonding orbitals in the absence of a distal imidazole. There is bonding overlap between Fe d_π and NO π* orbitals, and an antibonding interaction with the imidazole lone pair, that inhibits backbonding. These orbitals account for the negative slope of the ν_FeN/ν_NO correlation. The orbital on the right, HOMO-6, describes an imidazole lone-pair interaction, mainly with the NO π* orbital, with some involvement of π orbitals on the porphine, on the side opposite the distal imidazole. The imidazole lone pair pushes electrons out of the NO π* orbital onto the porphine, thereby strengthening the N–O bond, with little effect on the Fe^III–NO bond. This interaction is the likely origin of the shift of the ν_FeN/ν_NO correlation to higher ν_NO when the distal imidazole interacts mainly with the O atom of NO.

Figure 10.

Selected HOMOs of the (ImH)N:→(NO)Fe^III(ImH) porphine model (C_s).

We note that the ν_FeN/ν_NO point for distal imidazole interacting with the NO₂-substituted porphine deviates from the general pattern, falling distinctly below the line for the other models (Figure 8). This deviation for the most electron-withdrawing substituent can be traced to HOMO-6 acquiring more Fe–NO antibonding character (data not shown) when the substituents are –NO₂, and moves above the backbonding orbitals, thereby governing the Fe^III–N–O bonding.

(c) O- Versus N-Directed Imidazole and Fe^III–N–O Bending

In the course of computing vibrational frequencies, we noted that the C_s distal imidazole model yielded a negative frequency, ~ −16 cm⁻¹, corresponding to rotation of the distal imidazole ring out of the plane to which it was constrained. While the negative frequency was low enough to preclude significant errors in the computed ν_NO and ν_FeN values, it did signal that the C_s constraint imposed an energy penalty, probably due to non-bonded contacts between the imidazole CH and the porphine macrocycle. At the close distance required for the N: lone pair interaction, this contact is within the sum of van der Waals radii (~ 3 Å).

When the C_s constraint was removed, the distal imidazole rotated out of the plane, as expected. In addition, however, it shifted toward the N atom of the NO (Table 3 and Figure 9b). The imidazole N: was now only slightly closer to the O_NO than the N_NO atom, 2.92 vs. 3.12 Å. Moreover, electron-withdrawing substituents pulled the (ImH)N: closer to the N_NO than the O_NO atom. They also induced significant Fe^III–N–O bending. In the extreme case of eight –NO₂ substituents, the distances of (ImH)N: to O_NO and N_NO were 2.80 and 2.53 Å, respectively, while the Fe^III–N–O angle was reduced to 156° (Table 3).

When the computed frequencies were placed on the ν_FeN/ν_NO plot (Figure 11), a surprising pattern emerged. The data for the more electron-donating β-substituents (–NH₂, –CH₃, –H) fall on a shifted backbonding correlation (Figure 11, top red line), similar to that found for the C_s model (Figure 11, blue line), but the data for the more electron-withdrawing β-substituents (–F, –Cl, –NO₂) fall on a positive correlation, both ν_NO and ν_FeN decreasing as the substituents became more electron-withdrawing (Figure 11, bottom red line). The break between the two correlations occurs between eight –H substituents (porphine) and four each of –F and –H substituents (H- and F₄-labeled points in Figure 11). With more electron-donating β-substituents the (ImH)N: stays farther from the N_NO than the O_NO atom, and the Fe^III–N–O unit stays nearly linear, while with more electron-withdrawing β-substituents the (ImH)N: moves closer to the N_NO than the O_NO atom, and the Fe^III–N–O unit bends strongly (Table 3).

Figure 11.

ν_FeN/ν_NO correlation plots computed for the (ImH)Fe^III(NO)PX₈ and (ImH)Fe^III(NO)PH₄F₄ complexes, without (blue points, data from Figure 8) and with a distal ImH allowed to relax to C₁ symmetry (red triangles). The black line is the positive correlation of ν_FeN and ν_NO, obtained by constraining the Fe^III–N–O angle at the indicated values in the (ImH)Fe^III(NO)P model. Also shown is the HOMO-1 a_2u-type orbital for the (ImH)Fe^III(NO)P model with the ∠Fe^III–N–O = 140° (electron density has been removed from the front side of the molecule, to show the Fe–N–O contribution).

A similar effect of the distal (ImH)N:→N_NO interaction was observed by constraining the distance between the distal ImH and N_NO at fixed values for unsubstituted Fe^III(NO) porphine (C₁ model). The closer the approach of the distal ImH to the N atom of NO, the more bent the Fe^III–N–O angle became, reaching 154° at a 2.4 Å distance. The Fe^III–NO and N–O bond lengths increased in concert as the distal (ImH)N:→N_NO distance decreased (Table S1, Figure S2). We note that the energy well for the interaction is shallow; only 0.4 kcal/mol are lost on increasing the N_NO···N_Im^d distance from the bottom of the well, 3.1 Å, to 3.5 Å. When the distal ImH was constrained to interact with O_NO (C_s model), the Fe^III–N–O unit remained nearly linear (Table S2, Figure S3).

The positive ν_FeN/ν_NO correlation implies an effect that overrides backbonding, and that correlates with Fe^III–N–O bending. To test the role of bending per se, we examined its effect on the porphine model, (ImH)Fe^III(NO)P, by constraining the Fe^III–N–O angle to different values (Table 3). Bending Fe^III–N–O reduced both ν_NO and ν_FeN, producing a positive correlation (Figure 11, black line); a similar effect of bending was reported by Linder and Rodgers.¹⁰ Indeed, the slope of the bending correlation is the same as that produced by increasingly electron-withdrawing β-substituents (Figure 11, bottom red line). We note that a positive correlation was earlier computed for the Fe^II–CO adduct, (ImH)Fe^II(CO)P, upon bending the Fe^II–C–O unit.⁵⁷ In both cases bending weakens the Fe–N(C) and N(C)–O bonds simultaneously.

The reason for this mutual weakening can be seen in the composition of the HOMO-1 molecular orbital, shown in Figure 11 (inset) for (ImH)Fe^III(NO)P when Fe^III–N–O is bent. Bending induces an antibonding combination of the d_xz and π*_NO orbitals, a feature noted earlier by Linder and Rodgers.¹⁰ It is this antibonding combination that produces simultaneous Fe^III–NO and N–O bond weakening, and a positive ν_FeN/ν_NO correlation.

In addition, bending induces interaction of π*_NO with the porphyrin a_2u-type orbital (Figure 11, inset). The a_2u orbital is concentrated on the pyrrole N atoms and the meso-C atoms, and is strongly affected by meso-substituents. As these substituents become electron-donating, the a_2u-π*_NO interaction becomes more favorable, inducing more Fe^III–N–O bending, thereby accounting for the positive ν_FeN/ν_NO correlation computed by Linder and Rogers when the meso-substituents were varied. However, the a_2u orbital has no amplitude at the β-carbon atoms of the pyrrole rings. Consequently, β-substituents do not induce Fe^III–N–O bending. Instead their inductive effect shifts the extent of Fe^III–NO backbonding, producing a negative ν_FeN/ν_NO correlation.

(d) Distal Water Interaction

Models were also constructed in which the distal imidazole was replaced by water molecules, in order to investigate a possible distal water effect on ν_FeN/ν_NO values of the nitrosyl ferriheme protein nitrophorin 4 (see below). Stable structures were found with the water O atoms pointing at the bound NO, somewhat closer to the O_NO than the N_NO atom (Table 3), with the Fe^III–N–O unit remaining nearly linear. As with distal imidazole lone pair interacting with O_NO, the effect of the interaction was to decrease ν_FeN but increase ν_NO, shifting the ν_FeN/ν_NO point above the interaction-free backbonding line. The H₂O effect was qualitatively similar, placing the frequency point on the correlation line formed by the models with distal (ImH)N:→O_NO interaction (Figure 8).

We note that Linder and Rodgers also found a stable structure for a formaldehyde molecule interacting with the bound NO in (CH₃S⁻)Fe^III(NO)P, modeling the interaction of a peptide carbonyl in the enzyme cytochrome P450 nitric oxide reductase (P450_nor).⁵⁸ As the contact distance was decreased, the computed Fe^III–NO and N–O bond distances both decreased slightly, suggesting a positive ν_FeN/ν_NO correlation with variable interaction strength. This effect had been observed earlier by Scheidt and coworkers in crystal polymorphs of [Fe^III(OEP)NO]ClO₄ (OEP = octaethylporphyrin) having the perchlorate counterion at different distances from the NO.^11,59 Both ν_FeN and ν_NO increased for the closer counterion contact.

Discussion

Fe^III(NO) Backbonding and Bending

Our RR spectroscopic and DFT modeling results establish that, in the absence of forces that induce bending, backbonding dominates the behavior of Fe^III(NO) adducts, consistent with the Fe^II(NO⁺) formulation of its electronic structure. A negative ν_FeN/ν_NO correlation is found, experimentally and computationally, for NO adducts of imidazole-ligated Fe(III) heme. The slope of the correlation is essentially the same whether the backbonding is varied by electron donating or withdrawing porphyrin β-substituents (computation, slope = −1.1), or by the protein electrostatic field (experiment, slope = −1.2). In addition, lone pair donation from a distal imidazole shifts the correlation to higher ν_NO values. This effect, seen experimentally in Mb^III(NO) variants, is accurately modeled via DFT, and is attributable to a specific orbital interaction that strengthens the N–O bond, with little effect on the Fe–NO bond.

However, Fe^III(NO) bending overrides the backbonding anti-correlation, and produces a positive ν_FeN/ν_NO correlation as the Fe–N–O angle is varied. This effect arises from induction of an antibonding combination of d_xz and π*_NO orbitals in HOMO-1 as Fe^III(NO) bends, resulting in simultaneous weakening of the Fe–NO and N–O bonds. Bending can result from a distal lone pair donor that is drawn close enough to interact with the N instead of the O atom of the NO.

There are interesting parallels between distal lone pair interactions with Fe^III(NO) adducts and distal H-bond interactions with Fe^II(NO) adducts, investigated earlier,^36,55 which can best be appreciated with the aid of the valence bond diagrams in Figure 12. In both cases, interaction at the O_NO atom polarizes the π bonds. Backbonding is enhanced by H-bonding to Fe^II(NO), but diminished by lone pair donation to Fe^III(NO), which, in addition strengthens the N–O bond via a specific orbital interaction. In both Fe oxidation states, interaction at the N_NO atom pulls electron density into the sp²-like antibonding orbital, inducing bending and weakening both the Fe–NO and N–O bonds.

Figure 12.

Different mechanisms of bond strength modulation through interaction with a distal imidazole lone pair (Fe^III–NO adducts) or a distal imidazole H-bond (Fe^II–NO adducts). A. Modulation through the O_NO polarization (backbonding). B. Modulation through interaction with N_NO (Fe–N–O bending).

For Fe^II(NO), DFT computation reveals stable H-bond interactions at either O_NO or N_NO, depending on the disposition of the H-bond donor.^55,56 Examples of both interactions are found in different heme proteins (see Ref. 55 and references therein). For Fe^III(NO), the preferred lone-pair interaction is with O_NO, unless the donor is drawn toward the N_NO by electron-withdrawing heme β-substituents, or unless the donor is a strong nucleophile, like a hydroxide ion (see below). Since protoporphyrin IX does not have electron-withdrawing substituents, and since protein side-chains are not strong nucleophiles, the Fe^III(NO) structural and vibrational data from ferriheme nitrosyls can be interpreted in terms of backbonding and lone-pair interactions with O_NO, provided the proximal ligand is histidine.

Implications for Imidazole-Ligated Ferriheme Nitrosyls

With the ν_FeN/ν_NO correlations established for Mb^III(NO), we can interpret reported vibrational data for Fe^III(NO) adducts of other imidazole-ligated proteins and models (Figure 13). The extensively characterized nitrophorins from the saliva of the ‘kissing’ bug Rhodnius prolixus are of particular interest, because of the stability of their Fe^III(NO) adducts. Two sets of Fe^III(NO) stretching frequencies have been reported for nitrophorin 4 (NP4), at pH 7.5 and 5.5.²² Intriguingly, these fall on the two lines described by Mb^III(NO) variants, with and without distal imidazole interactions. At the lower pH, NP4 is in a ‘closed’ form in which hydrophobic residues are packed around the bound NO. There is no potential lone pair donor, and the ν_FeN/ν_NO (590/1904 cm⁻¹) position, at the low end of the imidazole-less backbonding line (Figure 13, red line), indicates an environment of low polarity.

Figure 13.

ν_FeN/ν_NO data for the indicated His-ligated Fe^III(NO) heme proteins, which fall on the plus (blue line) or minus (red line) distal His lines of Mb^III(NO) variants shown in Figure 6 or form a parallel intermediate linear correlation (dashed line with r = 0.96 and slope = −1.0). NOR = nitric oxide reductase, hHO-1 = human heme oxygenase-1, HbA = human hemoglobin, HbN = truncated hemoglobin with distal Tyr and Gln residues, NP4 = nitrophorin 4 and NP1 = nitrophorin 1 at the indicated pH values, HRP = horseradish peroxidase, and H64L/V68N = His64Leu/Val68Asn Mb double mutant. Pyridine adduct of a model complex Fe^IIIOEP(NO) (OEP = octaethylporphyrin) is also shown ().

An early crystal structure of the Fe^III(NO) adduct of NP4 at pH 5.6 was refined as two conformers, one with a linear and one with a severely bent (110°) Fe–N–O unit.⁶⁰ The severely bent conformation was later attributed to partial reduction to Fe^II(NO) in the X-ray beam, and a higher resolution gave a single conformer with ∠ Fe–N–O = 156°, but with thermal disorder at the O position.⁶¹ It seems likely that this determination too was influenced by X-ray-induced reduction, since the ν_FeN/ν_NO data would have shown a pronounced deviation below the backbonding line for this degree of bending (see Figure 11), and this is not observed.

At pH 7.5, a conformational change in NP4 induces disorder in a pair of loops lining the bound NO and opens the distal heme pocket to solvent. The ν_FeN/ν_NO data (593/1921 cm⁻¹) place this form on the distal lone-pair line (Figure 13, blue line), yet there is no distal histidine, nor any other potential donor side-chain on the distal side of the heme; an aspartate residue is ~6 Å from the Fe,⁶¹ too far to donate an electron pair, though it could polarize an intervening water molecule. The actual ferriheme–NO structure is not available at pH 7.5, because of instability to X-ray-induced reduction, but crystal structures of NH₃ and CN⁻ adducts reveal an open, water-filled distal pocket.^60–62 In the CN⁻ adduct, a distal threonine residue orients a water molecule, holding it 2.5 Å from CN⁻ (pdb ID: 1EQD). Our DFT modeling indicates that a lone-pair donating water molecule could have an effect on ν_FeN/ν_NO equivalent to a distal imidazole (Table 3 and Figure 8).

Like NP4, nitrophorin 1 (NP1) has an open binding pocket at elevated pH, but the crystal structure of the NP1(CN⁻) adduct (pdb ID: 3NP1) shows fewer water molecules than in NP4(CN⁻), and none are localized to the immediate vicinity of the CN⁻, in contrast to NP4(CN⁻) (pdb ID: 1EQD). Consistent with this difference, the ν_FeN/ν_NO data point for NP1(NO) at pH 8.0^20,21 is shifted to an intermediate position (Figure 13, dashed line), indicating a weaker lone-pair donor interaction.

Human heme oxygenase-1 (hHO-1)¹⁹ falls on the distal imidazole line (Figure 13, blue line), but in this case the lone pair donor is the carbonyl oxygen of distal Gly139, which the crystal structure shows to be in close contact with bound NO, as seen in the Fe^II(NO) adduct (3.1 and 2.7 Å distances to N_NO and O_NO, respectively).⁶³ Reported data for the pyridine adduct of Fe^IIIOEP(NO)^25,26 also fall on the distal imidazole line, indicating that excess pyridine, present to ensure adduct formation, provided a lone pair interaction.

In nitric oxide reductase (NOR) from Paracoccus denitrificans, a cytochrome bc protein, NO binds to a vacant coordination site on imidazole-ligated ferriheme b₃.^64,65 Spectroscopic studies indicate that a trisimidazole-ligated non-heme Fe_B is distal to the heme b₃ in NOR,⁶⁶ analogous to the Cu_B center in cytochrome c oxidase (CcO),^67–69 and is likely involved in NO reduction. A model complex has been synthesized, and interestingly, its Fe^III(NO) vibrational frequencies,⁷⁰ ν_FeN/ν_NO = 610/1924 cm⁻¹, place it far above the Mb^III(NO) correlations in Figure 13. This placement is reminiscent of the elevated ν_FeC/ν_CO frequencies found for the CO adduct of CcO.^71,72 In that case, a compression effect of the heme/Cu_B dinuclear site was invoked to explain the elevated ν_FeC,⁴³ and a similar explanation may apply to the NOR model. However, NOR itself does not show this effect. Its Fe^III(NO) adduct (ν_FeN/ν_NO = 594/1904 cm⁻¹)⁷ falls on the imidazole-less line (Figure 13, red line), indicating that the protein site is less constrained than the model.

In addition to NP1, several other proteins occupy intermediate positions in Figure 13; a dashed line has been drawn through them to guide the eye. One of them is the Mb variant H64L/V68N, in which the distal His is absent, but a new potential donor is introduced in place of Val68, at the side of the bound NO (Figure 2). This orientation may limit lone pair donation by the carbonyl O^ε atom of the Asn68 side-chain, accounting for the intermediate ν_FeN/ν_NO point. However, Asn68 could also increase positive polarity of the distal environment via its N^εH atoms, thereby enhancing backbonding, as reflected in the elevated ν_FeN. Backbonding is even more pronounced for the Fe^III(NO) adduct of horseradish peroxidase (HRP), which has unusually high ν_FeN (Figure 13), while having an intermediate ν_NO. Distal Arg, as well as His, residues enhance positive polarity in HRP. Meanwhile, the distal His (or an intervening water molecule, as judged from the crystal structure of the Fe^II(CO) adduct, pdb ID: 1ATJ) can provide lone pair donation. However, the ν_NO value may in this case be limited by the anionic character of the strongly H-bonded proximal His ligand. This effect was also seen in the Fe^II(CO) adduct of HRP, for which the ν_FeC/ν_CO point falls below the neutral His ligand backbonding line.⁴³

Energetics of the Lone Pair Interaction

The switch from H-bond donation by a distal His for Fe^II(NO) to lone pair donation for Fe^III(NO) is a natural consequence of the increase in positive charge on the bound NO ligand. The lone pair interaction stabilizes the Fe^III(NO) unit modestly, as can be seen from the changes in the NO association rate (k_on), dissociation rate (k_off), and association equilibrium (K_NO) constants, shown in Table 4. Mutation of His64 to apolar amino acids results in a ~7 to 10-fold increase in the rate constant for NO dissociation, k_off, from the Fe^III(NO) adduct of Mb (Table 4).73 Assuming that the slowing of NO dissociation by His64 reflects the lone pair stabilization energy, the factor of 7–10 translates to stabilization of bound NO in native ferric Mb by −1.0 to −1.3 kcal/mol (i.e., –RTln(k_off,mutant/k_off,wt)).³⁹ This estimate agrees well with our DFT-computed energy released upon moving a distal imidazole from a non-bonded distance of 3.4 Å to its optimum distance, 2.8 Å, ~ −1.2 kcal/mol for the C_s model (which does not induce bending, Figure S3). This interpretation is supported by the results for the His64 to Gln64 mutation in which lone pair donation from the O^ε atom is possible, perhaps accounting for the smaller 5-fold increase in k_off for this variant (Table 4).

Table 4.

Kinetic Parameters of NO Binding to Ferric and Ferrous States of WT Mb and Its Mutants.^a

Mb	kon (μM−1s−1)	koff (s−1)	KNO (μM−1)	Ref.
MbIII(NO)
WT	0.08	12	0.0067	73, 75
His64Gln	8.2	59	0.19	73, 75
His64Leu	44	80	0.55	73, 75
His64Val	160	120	1.3	73
MbII(NO)
WT	22	0.00010	220,000	39, 73, 74
His64Gln	43	0.00011	370,000	39, 74
His64Leu	190	0.00013	1,500,000	39, 74
His64Val	270	0.0011	250,000	73

^ak_on = NO association rate constant, k_off = NO dissociation rate constant, K_NO = NO affinity constant (k_on/k_off).

The NO affinity of Fe^IIIMb is markedly increased ~30, 80, and 200-fold for the Gln64, Leu64, and Val64 variants, respectively, because the NO association rate constants, k_on, increase ~100 to 2,000-fold (Table 4).⁷³ The slower bimolecular rates for NO binding to native Fe^IIIMb reflect the free energy barrier for displacement of the water molecule coordinated to the iron atom, which in wild-type Fe^IIIMb is also stabilized by lone pair donation by His64. As described by Olson and Phillips,³⁹ the unfavorable free energy required to disrupt water coordination can be estimated from RTln(k_on,mutant/k_on,wt) and the value is ~ +3 to +4 kcal/mol, which is substantially larger than the favorable Fe^III(NO) lone-pair interaction energy that stabilizes NO coordinated to Fe^IIIMb. The net result is that the affinity of wild-type Fe^IIIMb for NO is very small compared to that for the apolar Val64 or Leu64 Fe^IIIMb mutants (Table 4). The same His64 mutations also increase the association rate constant for NO binding to Fe^IIMb, but in this case, the increases in k_on for Fe^IIMb are much smaller (10 to 20-fold) than those for Fe^IIIMb (500 to 1000-fold, Table 4). The favorable free energy associated with proton donation by His64 to NO bound to Fe^IIMb, estimated from the ratio of the mutant versus wt NO dissociation rate constants (≤ ~1 kcal/mol) is equal to or only slightly smaller than the unfavorable free energy associated with displacement of non-coordinated, internal water H-bonded to the distal histidine in Fe^IIMb, which can be estimated from the ratio of the corresponding association constants (~ +1.3 kcal/mol). The net result is that, unlike the situation for Fe^IIIMb, only small changes in NO affinity for Fe^IIMb are observed for His64 mutations (Table 4).³⁹

Fe^III(NO) Bending by Nucleophillic Attack: Reductive Nitrosylation and NO Reduction

Because electron-withdrawing substituents are absent in protoporphyrin IX, a distal imidazole is attracted mainly to the O_NO atom of Fe^III(NO) in ferriheme proteins. This is true also of water molecules (see above) or peptide carbonyl groups.⁵⁸ However, it is likely that stronger nucleophiles would shift to the N_NO atom, bending the Fe^III–N–O unit and interacting with the sp² orbital.

Nucleophile-induced bending is the likely pathway for “reductive nitrosylation”,^76–78 in which hydroxide ion attacks the bound NO, forming nitrous acid, HONO, and Fe(II) heme (Figure 14). The Fe(II) heme is trapped by coordination with excess NO. Consistent with this mechanism, the rate of Fe^III(NO) autoreduction increases rapidly with increasing pH.⁷⁸ To model this pathway, we carried out DFT optimization of (ImH)Fe^III(NO)P with hydroxide ion placed in proximity to the bound NO. Formation of a HONO intermediate has been proposed, based on kinetic studies of reductive nitrosylation in buffered solutions of Mb^III, Hb^III and Cyt^III.⁷⁶ Consistent with this proposal, the modeled HO⁻ promptly attacked the N_NO atom, causing bending of the Fe^III(NO) unit and formation of a stable HONO adduct (Figure 14, inset). This adduct is characterized by elongated Fe–N_HONO and Fe–N_Im^p bonds, and a lowered Mulliken charge on Fe (1.071e vs. 1.106e for the starting Fe^III(NO) porphine), consistent with Fe reduction. In principle, a similar mechanism of reductive nitrosylation could be available to other nucleophiles, including nitrite, thiols, and amines.

Figure 14.

Proposed pathway for reductive nitrosylation of Fe^III(NO)P. Inset: DFT-optimized structure of the adduct obtained as a result of interaction between HO⁻ and (ImH)Fe^III(NO)P. Indicated are bond distances (Å, black), and Mulliken charges (blue).

We speculate that NO reduction to N₂O might proceed by a similar pathway, with bound NO⁻ providing the nucleophile for N–N bond formation. Collman and coworkers have shown that their model compound for the binuclear nitric oxide reductase produces N₂O and a Fe^III–O–Fe^III adduct upon treatment of the di-ferrous compound with NO.⁷⁰ They were able to detect intermediates with NO bound successively to the non-porphyrin and the porphyrin Fe(II),⁷⁹ providing evidence for a ‘trans’ mechanism, in which the reacting NO molecules reside transiently on both Fe atoms of the binuclear complex. The reaction pathway for this di-NO adduct is uncertain, but might involve electron transfer from the heme to the non-heme Fe(II), producing Fe_B^I(NO) at the non-heme Fe_B site, capable of nucleophillic attack on the heme Fe^III(NO) (Figure 15a). The driving force would be stabilization of the Fe(III) by the porphyrinate di-anion, and of the Fe(I) by the neutral histidine ligands at the non-heme site. The much higher Fe(III/II) reduction potential of the heme than the non-heme site, 0.32 vs. 0.06 V,⁷⁰ is consistent with this site preference. Alternatively (Figure 15b), the first step in the pathway could be coupling of the two Fe(II)-bound NO neutral radicals, accompanied by concerted electron transfer from the two Fe(II)’s. However, spin restriction and the asymmetry of the binuclear site argue against this symmetric mechanism. The choice between the two pathways will rest upon detailed computation of the energy landscape for electron reorganization. However, the energy landscape may differ somewhat between the model compound and the NOR enzyme, since, as noted above, the model and the protein differ significantly in their Fe^III(NO) vibrational frequencies, in a manner suggesting greater rigidity in the model.

Figure 15.

Alternative pathways for N₂O production from the di-ferrous complex of the NOR binuclear model compund.⁷⁰,⁷⁹

Fe^III(NO) Bending by Thiolate Ligation: Cysteinyl Hemes

In addition to a lone pair interaction at the N_NO atom, Fe^III–N–O bending can be induced by mixing of the d_z² and d_xz orbitals, or more precisely the d_z²-σ_n and d_xz-π*_NO orbitals, as has been discussed by others.^80–83 This is the source of bending when thiolate is a ligand. The strongly donating RS⁻ drives up the d_z²-σ_n energy, and increases its mixing with d_xz-π*_NO. The crystal structure of a thiolate-ligated Fe^III(NO) adduct reveals a 160° Fe–N–O angle,⁸⁴ consistent with structures of cysteinyl heme proteins,^28,29,85 which also produce a positive ν_FeN/ν_NO correlation (Figure 1). For the model complex (CH₃S⁻)Fe^III(NO)P, Linder and Rodgers computed an Fe–N–O angle of 159°, which increased to 180° when the CH₃S⁻ was protonated, thereby lowering its donor strength.⁵⁸ Paulat and Lehnert computed structures and vibrational frequencies for Fe^III(NO)P, with thiolate ligands having zero, one and two H-bonds to the sulfur atom.⁸⁶ Both ν_FeN and ν_NO increased as the number of H-bonds increased, reflecting diminished d_z²-σ_n/d_xz-π*_NO mixing as the thiolate donor strength diminished. A similar effect underlies the cysteinyl heme protein ν_FeN/ν_NO correlation in Figure 1. The ν_FeN/ν_NO frequencies increase in the order cytochrome P450_cam < cytochrome P450_nor < chloroperoxidase (CPO), which is also the order of increased H-bonding to the cysteinyl sulfur, as judged by analyses of the crystal structures,⁵ and also by vibrational data on the corresponding CO adducts.⁴³

Conclusions

For imidazole-ligated Fe^III(NO) adducts ν_FeN and ν_NO are negatively correlated, consistent with variations in backbonding, but distal lone pair donors induce an additional increase in ν_NO due to a specific orbital interaction. These donors can include imidazole, but also peptide carbonyl groups and water molecules. The lone pair interaction of a distal imidazole represents approximately −1.0 to −1.3 kcal/mol of stabilization energy.

The ν_FeN and ν_NO frequencies become positively correlated when the Fe^III(NO) moiety bends, since the Fe^III–NO and N–O bonds are then weakened simultaneously. Bending can be induced by different stereoelectronic mechanisms, including:

1] electron-donating substituents at the heme meso-positions, which induce mixing of the porphyrin a_2u-type orbital with a bending-induced antibonding combination of d_xz and π*_NO orbitals,

2] a distal lone pair drawn close enough by electron-withdrawing β-substituents, or by strong nucleophilicity, to interact with a bending-induced sp²-like orbital on N_NO, and

3] a strongly donating proximal axial ligand that raises the d_z² orbital energy and induces mixing of the d_z²-σ_n and d_xz-π*_NO orbitals.

The first two of these mechanisms are generally inapplicable to heme proteins, since the protoporphyrin IX meso-substituents (–H) are not electron-donating and the β-substituents (methyl, vinyl, propionate) are not electron-withdrawing, and since the available lone pair donors are generally weak nucleophiles. However, attack of a strong nucleophile is facilitated by bending-induced exposure of the empty orbital on the N_NO atom. Examples include attack of hydroxide to form a transient HONO adduct, during reductive nitrosylation, and possibly the attack of Fe_B-bound NO⁻ in the NO coupling reaction induced by NO reductase.

The third mechanism is operative in heme proteins with the strongly donating cysteinate ligand. Here, the ν_FeN and ν_NO frequencies are positively correlated, with the degree of bending modulated by H-bond donation to the cysteinate.