A Combined NRVS and DFT Study of Fe^IV=O Model Complexes: a Diagnostic Method for the Elucidation of Non-Heme Iron Enzyme Intermediates

Abstract

Fe^IV=O biomimetic model complexes have been characterized using nuclear vibrational resonance spectroscopy (NRVS). Features and systematic trends in the low energy region reflect equatorial and axial bonding differences that relate to differences in reactivity. These trends have been computationally extended to predict the spectra of putative Fe^IV=O intermediates in non-heme iron enzymes and show the utility of the NRVS method for structural insight.

Fe^IV=O intermediates have been shown to be key catalytic species in a growing number of mononuclear non-heme iron (NHI) enzymes. Various combinations of Mößbauer (Mb), electron-nuclear double resonance, magnetic circular dichroism (MCD), and extended x-ray absorption data exist on these, providing electronic insight and bonding descriptions of the Fe^IV=O unit.[1] However information regarding other chemically important features necessary for mechanistic insight (overall structure) is mostly lacking. Resonance Raman (rR) and IR spectroscopy can, in principle, provide such insight, however, due to selection rules, low absorption intensities (NHI species lack the intense Soret band), and other factors, experimental data are limited.[2] Herein, we present vibrational data obtained by nuclear resonance vibrational spectroscopy (NRVS) on three Fe^IV=O complexes interpreted via coupling with DFT calculations, with particular focus on understanding the low energy spectral region. The data show systematic trends that provide molecular level insight reflecting the Fe^IV ligand environment and reactivity. Computationally, these studies are extended to models of the Fe^IV=O species in HmaS/HPPD (vide infra) and show that Fe^IV species can be experimentally differentiated using NRVS.

The NRVS experiment employs synchrotron radiation tuned to the 14.4 keV ⁵⁷Fe nuclear Mb transition.[3–5] Inelastic scattering, where phonon annihilation and creation events couple with nuclear excitation, yields the Fe partial vibrational density of states (PVDOS) NRVS spectrum. This is analogous to Stokes and anti-Stokes rR scattering, but NRVS intensity is gained through Fe displacement (ΔFe), via the mode composition factor, and provides selective enhancement of all Fe-core modes with no background.[6] This method has been applied to Fe-S and heme enzymes, nitrogenase and model complexes.[7–16] NRVS data for an Fe^V-nitrido complex with a ligand similar to 1 (vide infra) have been reported.[16] However, detailed analysis in the low energy region was complicated by inhomogeneity in the sample.[16]

Fe^IV=O (S = 1) functional models have been synthesized with 1,4,8,11-tetramethyl-1,4,8,11-tetraazacyclotetradecane (1); N,N-bis(2-pyridylmethyl)-N-bis(2-pyridyl)methylamine (2); and N-benzyl-N,N′,N′-tris(2-pyridylmethyl)-1,2-diaminoethane (3) ligands.[17–19] Each complex is 6-coordinate: 1 has an equatorial (eq) ring with tertiary amines and an axial (ax) NCCH₃ ligand; 2 has eq pyridines (py) and a tethered ax tertiary amine and 3 is quite rhombic with two tertiary amine and three py ligands. Structures are available for 1 and 2. Note, 2 and 3 are more reactive than 1 in their ability to H-atom abstract from cyclohexane.[20]

The NRVS data for 1, 2, and 3 are shown in Figure 1. Isotope sensitive peaks are observed at 831 (796 in ¹⁸O), 820 (788) and 824 (786) cm⁻¹, respectively, that can be assigned as ν₂, the FeO stretch (normal modes in Figure 2).[21, 22] Resolution on NRVS energies is 8 cm⁻¹. These correlate well with the calculated energies and isotope shifts, see SI and Figure S2. This mode has been observed only in 1 at 834 by IR and at 839 cm⁻¹ by rR, in good agreement with our results.[19, 23] The lowerν₂ frequencies for 2 and 3 relative to 1 imply that their FeO bonds are weaker. Excited state MCD data have shown the FeO π bond is stronger in 2 relative to 1 therefore the σ bond must be weaker.[24] This can be rationalized by considering the stronger donation of the eq ligands in 2 (vide infra) which donate into the donut of d_z2 and compete with the oxo.

Figure 1.

NRVS PVDOS for 1, 2 and 3 with ¹⁶O (solid lines) and ¹⁸O (dotted lines), insets show the 3D structures of complexes.

Figure 2.

Normal modes for C_4v complexes adapted from ref.[21] Modes with ΔFe are boxed. Nomenclature follows Nakamoto.[22]

Moving down in energy, a weak peak is observed at 653 cm⁻¹ in 2 and 3, while for 1 the next resolvable peak is at 526 cm⁻¹. In 2 and 3 this is assigned, based on DFT calculations, as ν₈, the asymmetric FeN_eq stretches. In the calculations, these are shifted to the 450 – 470 cm⁻¹ region in 1 and not resolved in the data. In total, 4 modes involve FeN_eq stretching (including the NRVS inactive ν₁ and ν₅) and probe eq bonding; the average energy of these sets is 355, 648 and 598 cm⁻¹, respectively. Calculations where the eq ring in 1 is cleaved (1′, see SI) show that the average FeN bond length shortens by 0.07 Å, the splitting of the 4 modes decreases and the average increases to 430 cm⁻¹. These results show that the chelate in 1 influences the eq bonding but that the overall strength is determined by the greater σ donation of the py ligands relative to tertiary amines which affects FeO bonding.

Continuing down in energy, the first intense feature is assigned as ν₃ (FeN_ax stretch) at 391, 367 and 389 cm⁻¹, in 1, 2 and 3, respectively. In the calculations these show variable intensity (bars in Figure 3) and mix with ν₁₁. The resulting ΔFe has both in-plane (ip) and out-of-plane (op) contributions from ν₁₁ and ν₃. The ip motion in 1 is restricted due to the chelate, reducing ΔFe, i.e. intensity. In 1′, ΔFe is restored and ν₃ shifts to 294 cm⁻¹, Figure S4. Calculations where the ax tether is removed in 2 (2′, see SI) show that ν₃ is shifted down in energy to 300 cm⁻¹_, consistent with a 0.05 Å increase of the FeN_ax bond, but the ip and op ΔFe are preserved. This mode reflects the bond trans to the oxo and constraints in the ligand environment.

Figure 3.

DFT calculated NRVS PVDOS for 1, 2 and 3 (¹⁶O (solid lines) and ¹⁸O (dotted lines)) with ΔFe included as bars.

The remaining two intense features are assigned as ν_11a/b (OFeN_ax bends), which split due to orthogonal differences in the eq ligand environment; ring size in 1, ax tether in 2 and eq chelate in 3. The splitting in 3 is 50 cm⁻¹ but only ~25 cm⁻¹ in 1 and 2 reflecting the contracted eq chelate between a pair of cis N’s in 3 and affords a probe of rhombic (perpendicular to the FeO) distortion. In 1′, the most symmetric model, the splitting is 6 cm⁻¹. In 2′ (where, the x and y axes are still inequivalent) the energy positions and splitting are unchanged but the calculated intensity is reduced due to increased N_ax motion. The splitting of ν₁₁ reflects the ligand environment perpendicular to the FeO bond and, combined with high intensity, should provide structural insight for Fe^IV=O intermediates.

(4-Hydroxyphenyl)pyruvate dioxygenase (HPPD) and (4-hydroxy)mandelate synthase (HmaS) are NHI enzymes containing the 2 His 1 Glu facial triad and use the same substrate (HPP) but exhibit different reaction mechanisms (electrophilic aromatic attack and H-atom abstraction, respectively) dictated by the orientation of HPP by the protein pocket.[25] The first step in both reactions is decarboxylation of HPP and computational studies on the resultant Fe^IV=O site show that two 5C geometries are reasonable: square pyramidal (sqp, 4) with the oxo in the eq plane and trigional bipyramidal (tbp) with the oxo axial (5).[25] Here we determine that NRVS can discriminate between these proposed structures. The calculated PVDOS are shown in Figure 4. The intense peaks at 301 and 424 cm⁻¹ in 4 are bends of the OFeN_trans unit (trans to the oxo) reminiscent of ν_11a/b, where the large splitting arises from loss of the cis ligand. Intense peaks are also at 237 and 837 cm⁻¹ which involve FeN_trans (~ν₃) and FeO (~ν₂) stretches. For 1, 2, and 3, ν₃ was observed at a higher energy relative to ν₁₁; the shift to lower energy in 4 is due to the increased effective mass of histidine. The NRVS data of 5 are distinctly different and can be understood considering the normal modes for a tbp complex, Figure S6. The degenerate ν₆ and ν₇ pairs involve trans ax bends and cis ML stretches. These are highly mixed in 5 and produce 4 peaks with similar intensity. The trans N_His stretch (~ν₃) is also intense at 257 cm⁻¹. In summary, with the exception of the FeO stretch, the intense modes occur at low energies and, due to the inverse energy dependency of NRVS intensity, are likely to be detectable in enzyme intermediates and afford a mechanism for structural characterization.

Figure 4.

DFT calculated NRVS PVDOS for 4 and 5 with ΔFe.

Analysis of the ground state NRVS vibrational data coupled with excited state data from previous MCD studies provides detailed bonding descriptions of 1, 2, and 3.[24, 26] The increased reactivity of 2 and 3 relative to 1 correlates with the strength of the FeO π bond (oxo p character in the π* LUMO, the frontier molecular orbital) and is determined by equatorial ligand σ donation. Additionally, the ν₃ andν₁₁ modes show systematic trends which reflect the ligand environment. Lower energy modes have generally not been used for geometric insight, but in NRVS spectra these are enhanced in intensity relative to higher energy stretches (e.g. characteristic metal ligand modes such as ν_FeO) and more likely to be detected in intermediates. This study uses these modes for electronic structural insight on Fe^IV=O models and computationally evaluates their utility for determination of active site structure in NHI enzyme intermediates.

Experimental Section

Samples prepared as previously described.[17, 18] NRVS was collected on beamline 3-ID at Advance Photon Source on multiple visits.[6, 7, 27] Spectra were collected from −25 and 125 meV in 0.25 meV steps. 5 to 12 scans were averaged and normalized from which the PVDOS was generated using the PHOENIX program.[28] Spin-unrestricted DFT calculations were performed using Gaussian 03 (see SI).