A comprehensive test set of epoxidation rate constants for iron(IV)-oxo porphyrin cation radical complexes

Article abstract of DOI:10.1039/c4sc02717e

Cytochrome P450 enzymes are heme based monoxygenases that catalyse a range of oxygen atom transfer reactions with various substrates, including aliphatic and aromatic hydroxylation as well as epoxidation reactions. The active species is short-lived and difficult to trap and characterize experimentally, moreover, it reacts in a regioselective manner with substrates leading to aliphatic hydroxylation and epoxidation products, but the origin of this regioselectivity is poorly understood. We have synthesized a model complex and studied it with low-pressure Fourier transform-ion cyclotron resonance (FT-ICR) mass spectrometry (MS). A novel approach was devised using the reaction of [Fe^III(TPFPP)]⁺ (TPFPP = meso-tetrakis(pentafluorophenyl)porphinato dianion) with iodosylbenzene as a terminal oxidant which leads to the production of ions corresponding to [Fe^IV(O)(TPFPP⁺a?￠)]⁺. This species was isolated in the gas-phase and studied in its reactivity with a variety of olefins. Product patterns and rate constants under Ideal Gas conditions were determined by FT-ICR MS. All substrates react with [Fe^IV(O)(TPFPP⁺a?￠)]⁺ by a more or less efficient oxygen atom transfer process. In addition, substrates with low ionization energies react by a charge-transfer channel, which enabled us to determine the electron affinity of [Fe^IV(O)(TPFPP⁺a?￠)]⁺ for the first time. Interestingly, no hydrogen atom abstraction pathways are observed for the reaction of [Fe^IV(O)(TPFPP⁺a?￠)]⁺ with prototypical olefins such as propene, cyclohexene and cyclohexadiene and also no kinetic isotope effect in the reaction rate is found, which suggests that the competition between epoxidation and hydroxylation - in the gas-phase - is in favour of substrate epoxidation. This notion further implies that P450 enzymes will need to adapt their substrate binding pocket, in order to enable favourable aliphatic hydroxylation over double bond epoxidation pathways. The MS studies yield a large test-set of experimental reaction rates of iron(iv)-oxo porphyrin cation radical complexes, so far unprecedented in the gas-phase, providing a benchmark for calibration studies using computational techniques. Preliminary computational results presented here confirm the observed trends excellently and rationalize the reactivities within the framework of thermochemical considerations and valence bond schemes.

Full text of DOI:10.1039/c4sc02717e

Chemical
Science
View Article Online
View Journal | View Issue
EDGE ARTICLE
A comprehensive test set of epoxidation rate
constants for iron(IV)–oxo porphyrin cation radical
complexes†
Cite this: Chem. Sci., 2015, 6, 1516
a
b
b
c
*
*
Mala A. Sainna, Suresh Kumar, Devesh Kumar, Simonetta Fornarini,
c
a
*
*
Maria Elisa Crestoni and Sam P. de Visser
Cytochrome P450 enzymes are heme based monoxygenases that catalyse a range of oxygen atom transfer
reactions with various substrates, including aliphatic and aromatic hydroxylation as well as epoxidation
reactions. The active species is short-lived and difficult to trap and characterize experimentally, moreover, it
reacts in a regioselective manner with substrates leading to aliphatic hydroxylation and epoxidation products,
but the origin of this regioselectivity is poorly understood. We have synthesized a model complex and studied
it with low-pressure Fourier transform-ion cyclotron resonance (FT-ICR) mass spectrometry (MS). A novel
approach was devised using the reaction of [FeIII(TPFPP)]+ (TPFPP ¼ meso-tetrakis(pentafluorophenyl)
porphinato dianion) with iodosylbenzene as a terminal oxidant which leads to the production of ions
corresponding to [FeIV(O)(TPFPP+c)]+. This species was isolated in the gas-phase and studied in its reactivity
with a variety of olefins. Product patterns and rate constants under Ideal Gas conditions were determined by
FT-ICR MS. All substrates react with [FeIV(O)(TPFPP+c)]+ by a more or less efficient oxygen atom transfer
process. In addition, substrates with low ionization energies react by a charge-transfer channel, which
enabled us to determine the electron affinity of [FeIV(O)(TPFPP+c)]+ for the first time. Interestingly, no
hydrogen atom abstraction pathways are observed for the reaction of [FeIV(O)(TPFPP+c)]+ with prototypical
olefins such as propene, cyclohexene and cyclohexadiene and also no kinetic isotope effect in the reaction
rate is found, which suggests that the competition between epoxidation and hydroxylation – in the gas-phase
– is in favour of substrate epoxidation. This notion further implies that P450 enzymes will need to adapt their
substrate binding pocket, in order to enable favourable aliphatic hydroxylation over double bond epoxidation
pathways. The MS studies yield a large test-set of experimental reaction rates of iron(IV)–oxo porphyrin cation
radical complexes, so far unprecedented in the gas-phase, providing a benchmark for calibration studies using
computational techniques. Preliminary computational results presented here confirm the observed trends
excellently and rationalize the reactivities within the framework of thermochemical considerations and
valence bond schemes.
Received 4th September 2014
Accepted 8th December 2014
DOI: 10.1039/c4sc02717e
www.rsc.org/chemicalscience
human health that include the biodegradation of xenobiotic
and drug molecules.1 Due to this broad chemical function the
Introduction
P450s can bind and activate a large range of substrates with
varying shapes and sizes. Generally, the P450s act as mon-
oxygenases, whereby they bind and utilize molecular oxygen
via a heme centre and transfer one of the oxygen atoms of O2
to a substrate, while the second oxygen atom leaves the
process as a water molecule. The P450s react with substrates
activating aliphatic and aromatic hydroxylation, epoxidation
and sulfoxidation reactions, but have also been reported to
catalyse desaturation and N-dealkylation reactions.2 There are
many different P450 isozymes and until early 2014 thousands
of different structures had been characterized.3 All P450s
share common features which include a catalytically active
heme group with a central iron atom that is linked to the
protein by the thiolate sulphur atom of a cysteinate side
chain.4
The cytochromes P450 are part of the body's natural defence
mechanism in the liver and perform vital functions for
aManchester Institute of Biotechnology and School of Chemical Engineering and
Analytical Science, The University of Manchester, 131 Princess Street, Manchester
M1 7DN, UK. E-mail: sam.devisser@manchester.ac.uk
bDepartment of Applied Physics, School for Physical Sciences, Babasaheb Bhimrao
Ambedkar University, Vidya Vihar, Rai Bareilly Road, Lucknow 226 025, India.
E-mail: dkclcre@yahoo.com
cDipartimento di Chimica
`
e Tecnologie del Farmaco, Universita di Roma “La
Sapienza”, P.le A. Moro 5, 00185, Roma, Italy. E-mail: mariaelisa.crestoni@
uniroma1.it; simonetta.fornarini@uniroma1.it
† Electronic supplementary information (ESI) available: Details of the
computational results, including Tables with group spin densities, charges,
absolute and relative energies as well as structures and Cartesian coordinates is
provided. See DOI: 10.1039/c4sc02717e
1516 | Chem. Sci., 2015, 6, 1516–1529
This journal is © The Royal Society of Chemistry 2015

View Article Online
Edge Article
Chemical Science
model substrates through oxygen atom transfer (OAT), electron
transfer (ET), hydride transfer and ligand addition. However, no
direct hydrogen atom transfer (HAT) with any tested substrate
took place.
In order to nd out what drives the OAT reaction of Cpd I
with olens, we decided to investigate the properties and
reactivities of [FeIV(O)(Por+c)]+ (Por ¼ porphine dianion) and
[FeIV(O)(TPFPP+c)]+ with FT-ICR MS and with density functional
theory (DFT) methods. These studies represent the rst
comprehensive – experimental and computational – study on
olen epoxidation by iron(IV)–oxo porphyrin cation radical
models and allow correlations to be established between the
OAT rate constant and the ionization energy (IE) of the olen.
These correlations are further supported and rationalized by
computational modelling.
Fig. 1 Active site of P450 as taken from the 2WM4 pdb file. Substrate
tyramine is highlighted in orange.
Methods
Materials
All chemicals used in the experiments, including (5,10,15,20-tetra-
kis(pentauorophenyl)porphinato)iron(III) chloride, [FeIII(TPFPP)]
Cl, were research grade products purchased from commercial
sources and used as received. All the solvents were analytical
grade. Ethylene, propylene, and E-2-butene were high purity
gases from Matheson Gas Products Inc. Iodosylbenzene
(C6H5IO) was prepared according to a published procedure and
Fig. 1 displays the structure of a typical P450 active site,
namely the one belonging to the CYP124 isozyme as taken from
the 2WM4 protein databank (pdb) le.5 As shown in Fig. 1 the
substrate (tyramine) is located in a cle nearby the heme, the
substrate binding pocket, which is in a tight orientation with
stabilizing hydrogen bonding interactions by several residues.
The vacant sixth coordination site of iron is the position where
molecular oxygen will bind during the catalytic cycle. The
process includes two reduction and two protonation steps to
synthesize the active species of P450 called Compound I (Cpd
I).6 Cpd I is highly reactive and therefore difficult to study
experimentally, however, a few reports on its spectroscopic
properties have appeared in the literature.7
Due to the short lifetime of Cpd I, studying catalytic mech-
anisms and reaction rates of P450 catalysed reactions is chal-
lenging;8 therefore, many studies have focused on biomimetic
iron–porphyrin complexes instead.9 These studies gave detailed
insight into the effect of axial and equatorial ligands,10 but also
on the local environment such as the substrate binding pocket.
A particularly useful method to establish the properties and
reactivity patterns of short-lived complexes, such as catalytic
intermediates, is Fourier transform – ion cyclotron resonance
(FT-ICR) mass spectrometry (MS).11 In FT-ICR MS, the charged
species of interest (either positive or negative ion) is trapped in a
collision cell for a specic time during which reactions with
neutral gases can occur and be studied at the prevailing low
pressure of the instrument. FT-ICR MS allows one to measure
the ion distributions and fragmentation patterns at varying
trapping time, thereby yielding insight into reactivities, and
enabling one to calculate rate constants and thermochemical
properties. In recent work, Crestoni, Fornarini and co-workers
have succeeded in trapping and characterizing the Cpd I
analogues of iron and manganese porphyrin complexes and
studied their reactivity with a selection of substrates.12 Thus, the
[MnV(O)(TPFPP)]+ complex (TPFPP ¼ meso-tetrakis (penta-
uorophenyl)porphinato dianion) was found to react with
stored at ꢀ20 C.13
ꢁ
Instrumental
All experiments were run on a Bruker BioApex Fourier trans-
form-ion cyclotron resonance (FT-ICR) mass spectrometer
equipped with an Apollo I electrospray ionization source, a 4.7 T
superconducting magnet, and a cylindrical innity cell. Analyte
solutions were continuously infused through a 50 mm internal
diameter fused-silica capillary at a ow rate of 120 mL hꢀ1 by a
syringe pump and ions were accumulated in a radiofrequency-
only hexapole ion guide for 0.8 s. The ion population, des-
olvated by a heated (380 K) N2 counter current drying gas, was
pulsed into the ICR cell at room temperature (300 K), where the
ion of interest was isolated by ion ejection procedures and
allowed to react with the selected neutral reagent (L) admitted
by a needle valve at stationary pressures in the range of 1.0–15 ꢂ
10ꢀ8 mbar. The pressure was measured with a cold-cathode
sensor (IKR Pfeiffer Balzers S.p.A., Milan, Italy) calibrated by
using the rate constant, kmethane ¼ 1.1 ꢂ 10ꢀ9 cm3 sꢀ1, for the
+
+
reference reaction CH4 c + CH4 / CH5 + CH3c and corrected
for different response factors.14
Pseudo-rst order rate constants were obtained from the
slope of the semi-logarithmic decrease of the parent ion abun-
dance as a function of time and divided by the substrate
concentration to determine the second-order rate constants
(kexp) at 300 K. The reaction efficiencies (F) are percentages of
the second-order rate constant with respect to the collision rate
constant (kADO), i.e. F ¼ kexp/kADO, calculated by the parame-
trized trajectory theory.15 These values and product ion
branching ratios were found to be independent of the pressure
This journal is © The Royal Society of Chemistry 2015
Chem. Sci., 2015, 6, 1516–1529 | 1517

View Article Online
Chemical Science
Edge Article
of the neutral and of the additional presence of an inert bath optimizations, analytical frequencies and geometry scans.
gas, i.e. argon. The reproducibility of the kexp values was within These studies explored the potential energy surfaces involving
10%, while the error of the absolute rate constants is estimated reactants, intermediates and products on different spin states
to be ꢃ30%.
in detail and generated starting structures for the transition
state optimizations. All local minima reported here had real
frequencies only and the transition states were characterized
by a single imaginary frequency for the correct mode. To
improve the energetics of these structures we did single point
calculations in the gas-phase with a triple-z quality basis set
on iron (LACV3P+) and 6-311+G* on the rest of the atoms,
basis set BS2. Subsequently, all geometries (local minima and
transition states) were reoptimized at the UB3LYP/BS2 and
UB3LYP-D3/BS2 levels of theory19,22 and characterized by an
analytical frequency analysis. Barrier heights reported in this
work were calculated relative to isolated reactants, although
using reactant complexes instead only minor changes are
observed (ESI†).
Sample preparation
The [FeIV(O)(TPFPP+c)]+ ion of interest was prepared by adding
iodosylbenzene (0.5 mM) to [FeIII(TPFPP)]Cl (10 mM) in a
CH3OH/CH2Cl2 (1 : 1) solvent mixture cooled at ꢀ40 ꢁC. The so-
formed [FeIV(O)(TPFPP+c)]+ ion persisted at this temperature for
about 1 h. The high-resolution electrospray ionization FT-ICR
mass analysis shows the presence of ions [FeIV(O)(TPFPP+c)]+ as
a prominent cluster centered at m/z 1044, along with the resting
form [FeIII(TPFPP)]+, characterized by the same isotopic pattern,
and centered at m/z 1028.
As already pointed out,16 the synthetic method yields a
portion of an isomeric species, likely oxidized on the porphyrin
ligand and ineffective in any OAT reaction to reductants. This
OAT-unreactive fraction of the overall ion population at m/z
1044 has been identied as a four-coordinate iron(III) complex
incorporating an O-atom on the porphyrin ligand, henceforth
denoted as [FeIII(TPFPP-O)]+.16 Its relative amount has been
evaluated exploiting the characteristic reaction with NO, dis-
playing different paths from the two isomeric species. In fact,
The effect of solvent on the rate constants was tested through
single point calculations using the self-consistent reactant eld
model as implemented in Gaussian with a dielectric constant
representing chloroform (3 ¼ 4.7113).
Ionization energies and bond dissociation energy (BDEOH
)
values were calculated as before23 and represent adiabatic
values for reaction 1 and 2, respectively and report UB3LYP/
BS2//UB3LYP/BS1 energies including ZPE and dispersion
corrections.
while [FeIII(TPFPP-O)]+ yields
a ligand addition product,
[FeIII(TPFPP-O)(NO)]+, the genuine iron(IV)–oxo complex
[FeIV(O)(TPFPP+c)]+ undergoes an OAT process by releasing
[FeIII(TPFPP)]+.
A / A+c + eꢀ + IEA
(1)
[FeIV(OH)(Por)]+ / [FeIV(O)(Por+c)]+ + Hc + BDEOH (2)
Computation
All calculations discussed here utilize density functional theory
(DFT) methods as implemented in the Jaguar and Gaussian-09
program packages.17 Two different models were investigated: (i)
[FeIV(O)(Por+c)]+ (A) that includes a porphyrin (Por) ring
with all side-chains abbreviated to hydrogen atoms, and (ii)
[FeIV(O)(TPFPP+c)]+ (B), Scheme 1. Similar to previous work of
ours in the eld,18 we use the unrestricted hybrid density
functional method UB3LYP19 as it was shown to reproduce the
kinetics of metal(iv)–oxo oxidants well.20 Initial exploratory
calculations employed a modest LANL2DZ basis set on iron and
6-31G on the rest of the atoms (basis set BS1)21 for geometry
Results
Gas phase reactivity with FT-ICR MS
Formation and characterization of naked [FeIV(O)(TPFPP+c)]+
ions. In an early survey, the preparation of a high-valent
iron(IV)–oxo porphyrin cation radical was achieved by controlled
oxidation of [FeIII(TPFPP)]Cl with H2O2 in a methanol solution.
This solution was then sampled by electrospray ionization
FT-ICR MS.12a However, under the selected experimental
conditions, in addition to the heterolytic cleavage of the
peroxide bond also a homolytic pathway is observed that
yields
a high-valent iron(IV)–hydroxo porphyrin complex
[FeIV(OH)(TPFPP)]+. Because the latter complex differs by one
mass unit only from the species of interest, this presence
complicates the ion distribution and the assignment of the
oxidant in the reaction mixture. Therefore, we aimed to
synthesize [FeIV(O)(TPFPP+c)]+ through an alternative, neater
mechanism.
A
methanol/dichloromethane solution of
[FeIII(TPFPP)]Cl was treated with iodosylbenzene (PhIO) as
oxygen atom donor. The reaction mixture, assayed by electro-
spray ionization and high resolution mass measurements by
FT-ICR mass spectrometry, displays an ion cluster centered at
m/z 1044.0116, which corresponds to the acquisition of just one
oxygen atom by the reactant species [FeIII(TPFPP)]+.
Scheme 1 Models investigated in this work.
1518 | Chem. Sci., 2015, 6, 1516–1529
This journal is © The Royal Society of Chemistry 2015

View Article Online
Edge Article
Chemical Science
The thus generated, naked ve-coordinate species proves to into the mass spectrometer as described above. Subsequently,
be resistant with respect to any unimolecular dissociation the [FeIV(O)(TPFPP+c)]+ ion was mass selected and trapped in
process (kdiss # 0.001 sꢀ1). However, when the ion-molecule the FT-ICR cell. The chosen olen (Sub) was present at constant
reactivity of the ion cluster at m/z 1044 is examined through the pressure in the cell and mass spectra were recorded at regular
addition of NO into the FT-ICR cell, two reaction products are time intervals aer isolation of the parent ion. Over time, all
observed: (i) a major fraction of ions reacts by OAT and forms [FeIV(O)(TPFPP+c)]+ was found to react and new peaks corre-
[FeIII(TPFPP)]+ and NO2 as the likely neutral product, (ii) a sponding to product ions appeared in the spectrum. Three
signicant percentage of the starting ions undergo a ligand different reaction paths were observed and monitored, namely
addition reaction to give a product with m/z 1074.
oxygen atom transfer (OAT), hydride transfer (HT) and charge
The product formed in the latter pathway has been sampled transfer (CT), as illustrated in Scheme 3.
by collision induced dissociation (CID), and displays the loss of
These three pathways lead to different product ions in the
NO as unique fragmentation channel thus suggesting that NO recorded spectra: (i) [FeIII(TPFPP)]+ ions are the products of the
has become part of the complex as intact ligand at a formerly OAT channel; (ii) ions at m/z corresponding to [Sub-H]+ are
vacant axial position. Note here that ligand addition is typical obtained from the HT processes; (iii) ions corresponding to
for the reactivity of tetracoordinate [FeIII(TPFPP)]+ ions.12a This Sub+c arise from a CT reaction. Interestingly, no evidence of
nding provides circumstantial evidence for an ion population hydrogen atom abstraction (HAT) is observed with any of the
of the same elemental composition as [FeIV(O)(TPFPP+c)]+, that selected olens listed in Table 1. In the case of two substrates,
may correspond to an isomeric structure oxidized on the cyclohexene and 1,4-cyclohexadiene, molecular addition (Add)
porphyrin ligand, designated as [FeIII(TPFPP-O)]+.
to the complex is also observed, leading to an ion with m/z value
The chemical titration reaction with NO is a way to resolve formally corresponding to [Fe(O)(TPFPP)(Sub)]+. It is possible
the isomeric distribution of electrosprayed ions at m/z that these addition complexes are products from a hydrogen
1044 and allowed us to establish the relative amount of atom abstraction reaction but currently this cannot be estab-
[FeIV(O)(TPFPP+c)]+ in each experiment, which is consistently lished from the product ions. However, experiments with fully
found to be equal to or larger than 60% of the total ion deuterated cyclohexene (cyclohexene-d10) versus that of cyclo-
population at m/z 1044.
hexene-h10 gave a rate constant ratio kH/kD very close to 1.
Reactivity of naked [FeIV(O)(TPFPP+c)]+ ions with olens. Our Reactions starting with an initial hydrogen atom abstraction
initial studies focused on experimentally determining the reaction normally encounter a large kinetic isotope effect (KIE ¼
reactivity of [FeIV(O)(TPFPP+c)]+ with the selected olens depic- kH/kD) of well greater than 1, hence these KIE experiments
ted in Scheme 2, and the results are summarized in Table 1. implicate that no rate determining hydrogen atom abstraction
These olens vary in molecular size and structure and their takes place here.
reported ionization energies span from 8.1 to 10.51 eV.24 The
Because the selected ions at m/z 1044 comprise isomeric
[FeIV(O)(TPFPP+c)]+ ion was prepared from the reaction of species as described in the previous paragraph, it is important
iodosylbenzene with [FeIII(TPFPP)]Cl in methanol/dichloro- to mention that in all experiments with the selected substrates
methane solution and transferred by electrospray ionization (olens and unsaturated hydrocarbons) an unreactive fraction
Scheme 2 Substrates investigated in this work.
This journal is © The Royal Society of Chemistry 2015
Chem. Sci., 2015, 6, 1516–1529 | 1519

View Article Online
Chemical Science
Edge Article
Table 1 Kinetic data and product distributions obtained for the gas phase reaction of [FeIV(O)(TPFPP+c)]+ with selected olefins as determined by
FT-ICR MS
IEa
kexp
kADO
Fd
HT
CT
OAT
Add
b
c
Substrate
Ethene
10.51
9.73
9.55
9.10
9.07
8.95
8.82
8.64
8.60
8.46
8.25
8.1–8.2
8.14
N/A
8.5 ꢂ 10ꢀ5
7.6 ꢂ 10ꢀ3
0.029
8.5
1 ꢂ 10ꢀ3
0.080
0.30
0.74
1.5
—
—
—
—
—
—
—
—
—
—
—
4
—
—
—
—
—
—
—
—
—
—
—
—
12
—
100
100
100
100
100
75
—
—
—
—
—
25
10
—
—
—
—
—
—
—
Propene
9.45
9.6
1-Butene
E-2-Butene
0.080
10.8
9.5
2,3-Dimethyl-1-butene
Cyclohexenee
1,4-Cyclohexadiene
2-Methoxy-1-propene
1,3-Pentadiene
Styrene
0.145
0.194–0.291
0.511
9.7
2–3
9.29
10.5
9.6
5.5
90
0.819
7.8
100
100
100
100
96
0.826
8.6
1.40
9.26
9.29
11.9
8.6
15
1,3-Cyclohexadiene
trans-b-Methylstyrene
Indene
1.58
17
2.97
25
3.18
37
2
86
b-Pinenef
4.32–4.7
9.4
46–50
—
100
a
Ionization energies (IE, eV) are from ref. 24. N/A stands for not available. b Second-order rate constants (kexp) in units of 10ꢀ10 cm3 moleculeꢀ1 sꢀ1
are measured at a temperature of 300 K in the FT-ICR cell. The estimated error in kexp is ꢃ30%, although the internal consistency of the data is
c
d
within ꢃ10%. Collision rate constants (kADO) evaluated with the parameterized trajectory theory. Reaction efficiency (%), F ¼ kexp/kADO
ꢂ
100. e The reaction with cyclohexene-d10 gave a rate constant within experimental error of that for cyclohexene-h10
. The IE for a-pinene is 8.07 eV.
f
is observed in a corresponding relative amount as the species porphyrin-oxidized complex and its contribution to the experi-
yielding the NO addition product in the probe reaction with ments has been thoroughly subtracted, an operation allowed by
nitric oxide. This species has been assigned the features of a the distinctly different (un)reactivity of the two isomeric
Scheme 3 Pathways observed for the reaction of [FeIV(O)(TPFPP+c)]+ ions (R ¼ C6F5) with selected substrates (Sub) as studied with FT-ICR MS.
1520 | Chem. Sci., 2015, 6, 1516–1529
This journal is © The Royal Society of Chemistry 2015

View Article Online
Edge Article
Chemical Science
complexes have been observed previously from other substrates
through oxygen atom-acceptor properties such as sulphides,
amines, and phosphorus containing complexes.12a,16
In addition, substrates with low ionization potential are
found to react via hydride transfer (trans-b-methylstyrene, and
indene) and/or charge transfer reactions (indene) as side-reac-
tions. We examined whether a charge-transfer (CT) reaction
occurs between [FeIV(O)(TPFPP+c)]+ and sampled substrates. As
in FT-ICR MS only exothermic reactions are able to occur due to
the low number of collisions in the gas, the data in Table 1 show
that the charge-transfer reaction is exothermic with indene as a
substrate (eqn (3)), but endothermic with other substrates.
Thus, the enthalpy change for the charge-transfer between
[FeIV(O)(TPFPP+c)]+ and indene should be equal to the differ-
ence between the ionization energy of the substrate and the
electron affinity (EA) of [FeIV(O)(TPFPP+c)]+. Since, no charge-
transfer reaction occurs with substrates with an ionization
energy above 8.2 eV (Table 1), this implies that the EA of gaseous
[FeIV(O)(TPFPP+c)]+ ions must be lower than 8.2 eV. Previous FT-
ICR MS studies predicted a lower limit of 7.5 eV for the EA of
[FeIV(O)(TPFPP+c)]+,16 which is in good general agreement with
the data in Table 1. DFT calculations on [FeIV(O)(Por+c)X] with
different axial ligands X have yielded electron affinities of 3.06
eV for X ¼ SHꢀ, whereas values of 6.41 and 6.12 eV are found for
systems with an imidazole or tyrosinate axial ligand, respec-
tively.26 Clearly, removal of the axial ligand is expected to raise
the EA of the iron(IV)–oxo porphyrin cation radical substantially
with respect to axially ligated systems due to the loss of inter-
action between the a2u molecular orbital with axial ligand
orbitals, vide infra.
Fig. 2 Time dependence of relative ion abundancies for the reaction
of [FeIV(O)(TPFPP+c)]+ (m/z 1044) with indene. Product ions are
[FeIII(TPFPP)]+ (m/z 1028), [Fe(TPFPP)(C9H8)O]+ (m/z 1160) and C9H8+c
(m/z 116). Experiments were performed in the presence of indene at
5.2 ꢂ 10ꢀ8 mbar in the FT-ICR cell.
complexes [FeIV(O)(TPFPP+c)]+ and [FeIII(TPFPP-O)]+. Hence-
forth, the implication of this species will not be further
discussed.
Fig. 2 gives an example of the time dependence of the relative
ion abundance of reactant and product ions as a function of
time for the reaction of [FeIV(O)(TPFPP+c)]+ with indene. In
general, the ion abundance of [FeIV(O)(TPFPP+c)]+ follows an
exponential decay as a function of time, as illustrated by its
reaction with indene in Fig. 2. The pseudo-rst order decay of
the reactant ion abundance as a function of time was then
converted into a bimolecular rate constant, kexp, which was
normalized by the respective collision rate constant (kADO) to
give the relative efficiency (F ¼ kexp/kADO ꢂ 100).15 The gas phase
reactivity of [FeIV(O)(TPFPP+c)]+ spans a wide range of reaction
efficiencies, with values varying from 0.001% to 0.08% for
terminal olens, like ethene and propene, up to ca. 50% for
electron-rich monoterpenes.
With most substrates a clean OAT conversion from reactants
to products is observed without production of by-products and
the time dependence shows a pattern like that displayed in
Fig. 2. In the case of cyclohexene, about 25% addition occurs
with the formation of an ion formally corresponding to
[Fe(TPFPP)(c-C6H10)O]+. The formation of a long-lived complex
with this substrate suggests that, within the metal coordination
sphere, the olen is activated to form an O-containing species
endowed with appreciable affinity for the metal. Ligand binding
to iron porphyrin complexes has been found to correlate with
the gas phase basicity of the ligand.25 It may thus be inferred
that the olen has turned into a more basic species, either an
epoxide by O-addition across the double bond or an alcohol by
allylic C–H bond activation. The nature of the adduct complex
has been further probed and conrmed by low-energy CID, and
reveals the formation of the reduced species [FeIII(TPFPP)]+
through the release of SubO. The so-formed complex from the
reaction of [FeIV(O)(TPFPP+c)]+ with the olen substrate may
then be depicted as [FeIII(TPFPP)(SubO)]+. Similar product
[FeIV(O)(TPFPP+c)]+ + indene /
[FeIV(O)(TPFPP)] + indene+c (3)
DHEq3 ¼ IEindene ꢀ EA[FeIV
< 0 kcal molꢀ1
(4)
(O)(TPFPP+)]
Interestingly, although cytochrome P450 isozymes react with
typical olens, such as propene, cyclohexene and Z- and E-
butene, to give a mixture of epoxide and enol products, actually
the precise product distributions depend on the specic P450
isozyme.27 Clearly, the substrate binding pocket and substrate
orientation within the enzyme play a key role in determining the
regioselectivity of the enzymatic reaction. In fact, enzymes
manage to control the regioselectivity of substrate activation
probably by binding the substrate under a specic orientation,
which raises the epoxidation barriers and/or lowers the
hydrogen atom abstraction barriers. Unfortunately, the FT-ICR
MS results are unable to unequivocally distinguish hydroxyl-
ation products from epoxidation products as both have the
same mass and are released as neutral molecules. As such, the
experiments do not provide direct evidence supporting olen
epoxidation over a C–H activation channel in the reactions of
[FeIV(O)(TPFPP+c)]+ with the selected olens. However, we are
able to show indirect evidence, vide infra. Computational
studies on the regioselectivity of propene epoxidation versus
hydroxylation by [FeIV(O)(Por+c)(SH)] gave lower epoxidation
barriers in the gas-phase reaction,28 but solvent and
This journal is © The Royal Society of Chemistry 2015
Chem. Sci., 2015, 6, 1516–1529 | 1521

View Article Online
Chemical Science
Edge Article
environmental effects reversed the ordering. The computational orbitals are dominated by the interactions of the metal 3d
studies from ref. 28, therefore, conrm the experimental trends orbitals with its ligands and several p-type porphyrin orbitals.
in Table 1. This nding underlines the benchmark role played Lowest in energy are a pair of s-type orbitals (sz2 and sxy): sz2
by a mechanistic study in the gas phase.
represents the s-interactions of the 3dz2 orbital on iron with the
Theoretically derived reaction paths, energetics and struc- 2pz orbital on oxygen, whereas the sxy gives the interactions of
tures. The experimental studies reported above present a the 3dxy orbital on iron with 2px,y orbitals on the four nitrogen
comprehensive test set of model reactions of iron(IV)–oxo atoms of the porphyrin ligand. The antibonding combinations
porphyrin cation radical complexes with olens for the rst of these two orbitals (s*z2 and sx*y) are high in energy and virtual.
time and enable extensive benchmarking and calibration of Also doubly occupied is the dx2ꢀy2 orbital, which is a lone-pair
computational methods and procedures against gas-phase orbital located in the plane of the porphyrin ring. Finally, the
(Ideal Gas conditions) rate constants. We decided to take the interaction of the metal 3dxz/3dyz with the 2px/2py on the oxygen
opportunity and calibrate previously used methods and proce- atom leads to a pair of pxz/pyz and a pair of p*xz/py*z set of
dures for DFT studies on these chemical systems and compare orbitals. The sz2, sxy, pxz and pyz bonding orbitals are doubly
to the results of the FT-ICR rates from Table 1. In addition, the occupied and low-lying in all calculations reported here. In
computational studies were performed to further understand addition to the metal-type orbitals there are also two porphyrin-
the substrate activation patterns by [FeIV(O)(TPFPP+c)]+ with type p-orbitals that in D4h symmetry have the labels a1u and a2u
.
olens, and rationalize the obtained trends. Before we discuss With a thiolate as axial ligand the a2u orbital strongly mixes with
details of the reaction mechanism and possible reactivity a 3pz orbital on sulphur and hence is destabilized in energy,29
trends, let us start with a detailed analysis of the reactant which strongly affects the electron affinity of the oxidant and
species, namely [FeIV(O)(Por+c)]+, 4,2A, and [FeIV(O)(TPFPP+c)]+, consequently is responsible for its push-effect.30
4,2B.
The set of orbitals displayed in Fig. 3 is occupied with 15
Fig. 3 displays the high-lying occupied and low-lying virtual electrons and as several of these orbitals are close in energy
orbitals of 4,2A; the orbitals for 4,2B look very similar. These there are a number of possibilities to distribute the electrons
Fig. 3 Molecular valence orbitals of 4A.
1522 | Chem. Sci., 2015, 6, 1516–1529
This journal is © The Royal Society of Chemistry 2015

View Article Online
Edge Article
Chemical Science
over the orbitals. In addition, states can also be found in various calculations done at UB3LYP/BS2 and UB3LYP-D3/BS2 give
spin states ranging from doublet to quartet and sextet, where we almost identical spin state orderings and relative energies,
identify the spin state with a superscript in front of the elec- which shows that dispersion is not a critical component for
4,2
4,2
tronic state label. Thus, the electronic state labelled as 4A2u has these chemical structures. Nevertheless, the
A
and
A
1u
2u
an overall quartet spin state and singly occupied a2u molecular states are close in energy and all four states could have a nite
2
2
2
2
orbital with overall electronic conguration: sz2 sxy pxz pyz
lifetime.
Optimized geometries of the 4,2A2u and 4,2A1u states are given
2
1
1
2
2
1
1
2
dx2ꢀy2 px*z p*yz a1u a2u1 or in short [core] dx2ꢀy2 px*z py*z a1u
a2u1. Similarly, we calculated the doublet spin state (2A2u state), in Fig. 4. Geometrically, no dramatic changes in bond lengths
where the unpaired electron in the a2u orbital is antiferromag- are obtained between the three optimization techniques. A
netically coupled to the unpaired electrons in the two p* small basis set gives slightly longer Fe–O distances than those
2
2
[
[
2
Y
orbitals: A2u ¼ [core] dx2ꢀy2 p*xz py*z a1u a2u
.
found with a triple-z basis set. The effect of dispersion is
Previous studies with either imidazole, acetonitrile or thio- negligible on the optimized geometries: UB3LYP/BS2 and
4,2
late as axial ligand,29,31 showed the
A
states to be close in UB3LYP-D3/BS2 give virtually the same chemical structures.
2u
energy and well below alternative states. However, this was due Addition of meso-substituents to the porphyrin ring such as
to considerable mixing of the a2u orbital with the axial ligand pentauorophenyl groups is not expected to dramatically
orbitals, which obviously is not possible in our chemical system change key bond lengths in the optimized geometries and
that lacks an axial ligand. However, in an isolated porphyrin relative energies of individual spin states.33 Thus, recent work of
macrocycle, the a1u and a2u orbitals are degenerate;32 therefore, the Goldberg group showed that meso-substituted manganese–
we decided to investigate a range of possible electronic states oxo porphyrinoid complexes retained the spin state ordering
for the pentacoordinated iron(IV)–oxo porphyrin cation radical and converged to a closed-shell singlet manganese(V)–oxo state
system, [FeIV(O)(Por+c)]+. Firstly, we tested the stability of the in all cases.33a
4,2
4,2
2
4,2
4,2
A
states and the alternative
A
states with [core] dx2ꢀy2
The calculations on the low-lying
A
and
A
states
1u
2u
1u
2u
p*xz py*z a1u a2u2 orbital occupation. In addition, we attempted reported in Table 2 and Fig. 4 show that geometrically there are
1
1
1
to generate models with the iron in oxidation state iron(V), i.e. very little differences between these states, but the spin state
2
1
2
2Pxz state with occupation [core] dx2ꢀy2 px*z a1u a2u2, or the ordering and relative energies are sensitive to the method and
iron in oxidation state iron(III), i.e. the 4A state with orbital basis set. Recent complete active site (CASSCF) and restricted
2
2
1
1
occupation [core] dx2ꢀy2 p*xz py*z a1u a2u1 and the 6Sxy,III state active site (RASSCF) calculations of Pierloot and co-workers34
2
1
1
1
1
with [core] dx2ꢀy2 p*xz p*yz s*xy a1u a2u1 occupation. However, calculated the 4,2A2u and 4,2A1u states of A within 1 kcal molꢀ1 of
all our attempts to calculate iron(III) or iron(V) states failed and each other with a small preference for the A1u states. However,
converged back to lower lying solutions with four electrons on they also located two low-lying iron(V) states, which we were
the metal in a formal iron(IV) oxidation state, hence the 2Pxz, 4A unable to characterize and for which no experimental evidence
6
and Sxy,III states are high in energy and inaccessible to our exist. Unfortunately, our chemical systems (in particular struc-
chemical system.
ture B) are too large to attempt calculations using the CASSCF
Table 2 summarizes relative energies of optimized geome- and RASSCF methods; therefore, we decided to continue with
tries of the various electronic spin states as calculated with UB3LYP instead.
different DFT methods for 2,4,6A. As follows from Table 2 all
Subsequently, we investigated the substrate epoxidation by
2
calculations of A give a A1u ground state that is nearly degen- [FeIV(O)(Por+c)]+, i.e. 4A. We nd the lowest lying barriers to
4
erate with the corresponding quartet spin state. In general, proceed from the A2u state and will focus on those in the
following. Previous studies on the epoxidation of olens by
[FeIV(O)(Por+c)(L)] with L ¼ NCCH3 or Clꢀ showed that the same
trends in reactivity are observed when the Por ligand is replaced
by TPFPP,35 hence the smaller model was used in this study. We
investigated substrate epoxidation with a range of olens:
Table 2 Relative energies of several low-lying electronic states of
[Fe(O)(Por+c)]+ (A)a
Ab
Ac
Ad
State
Conguration
DE + ZPE
DE + ZPE
DE + ZPE
2A1u
4A1u
2A2u
4A2u
6A2u
4Dxy
4Dzz
d2 p*xz p*yz a1u
0.00
0.71
1.65
1.25
9.25
9.69
19.53
0.00
0.19
3.75
3.42
18.68
ND
0.00
0.21
3.80
3.47
19.38
ND
[
[
Y
[
Y
[
d2 p*xz p*yz a1u
[
[
[
[
d2 p*xz p*yz a2u
d2 p*xz p*yz a2u
[
[
d[ p*xz p*yz sx*y a2u
[
[
[
[
Y
Y
d[ p*xz p*yz sx*y a1u
[
[
[
d[ p*xz p*yz sz*2 a2u
ND
ND
[
[
[
a
2
Relative energies in kcal molꢀ1 with respect to the A1u state, ND
b
stands for not determined. Energies obtained at UB3LYP/BS2//
UB3LYP/BS1 level of theory. Energies and geometries calculated at
UB3LYP/BS2 level of theory. Energies and geometries calculated at
c
4,2
4,2
Fig. 4 Optimized geometries of the
A
2u
and
A
states of 4,2A as
1u
d
calculated with UB3LYP/BS1 [UB3LYP/BS2] {UB3LYP-D3/BS2} with
UB3LYP-D3/BS2 level of theory.
Fe–O bond lengths in angstroms.
This journal is © The Royal Society of Chemistry 2015
Chem. Sci., 2015, 6, 1516–1529 | 1523

View Article Online
Chemical Science
Edge Article
ethene (1), propene (2), 1-butene (3), E-2-butene (4), cyclohexene resemblance to substrate epoxidation barriers calculated
(5), 1,3-cyclohexadiene (6), styrene (7), trans-b-methylstyrene (8), previously for P450 Cpd I reactions with olens.36 Electroni-
Z-2-butene (9) and 2-pinene (10). For all substrates we calcu- cally, all transition states are accomplished by single electron
lated the full potential energy prole from reactants to epoxide transfer from the substrate into the a2u orbital and the
products, see ESI,† but for space restrictions we will focus on formation of an [FeIV(OSub)(TPFPP)]+ transition state with
2
1
1
2
2
1
the rate determining C–O bond formation transition states orbital occupation [core] dx2ꢀy2 p*xz py*z a1u a2u fSub with
(TSCO) only. All reactions are concerted with a single C–O acti- fSub a radical on the substrate group.
vation barrier leading to epoxide product complexes PE. This is
unusual as previous calculations on substrate epoxidation by
Discussion
Cpd I models gave a stepwise mechanism via a radical inter-
mediate that via a ring-closure barrier was separated from
The present work gives a detailed and extensive overview on the
epoxide product complexes.36 The orientation of the substrate
reactivity of iron(IV)–oxo porphyrin cation radical systems with a
and the strong displacement of the metal from the porphyrin
plane are the likely reason for the fact that radical intermediates
are saddlepoints here. Thus, the ring-closure barrier on the
test-set of olens. We determined rate constants and measured
product ion distributions in the gas phase using FT-ICR MS.
This comprehensive set of transition metal containing reactiv-
quartet spin state surface involves an electron transfer from
ities is unique and will enable computation to benchmark and
substrate into s*z2. The latter orbital in iron–porphyrin
calibrate its methods effectively. This is particularly important
complexes with axial ligand, e.g. thiolate, contains a strong
for transition metal complexes, such as iron(IV)-oxo species,
contribution from axial ligand orbitals (3pz) and therefore is
where the reproducibility of the computational (DFT) methods
high in energy. Since our particular system lacks an axial ligand,
sometimes varies strongly depending on the density functional
the s*z2 orbital is considerably lower in energy and as a conse-
quence the lifetime of the radical intermediate is reduced and
the reaction to form products is now concerted.
Fig. 5 gives the optimized geometries of the C–O activation
transition states (TSCO) for all substrates. Generally, the tran-
sition states occur early with a long C–O distance and relatively
method used, the basis set, environmental perturbations,
dispersion effects etc.37 In this work, we supplemented the
experimental studies with a series of preliminary DFT calcula-
tions for two reasons: (i) to validate and calibrate computational
methods against experiment; (ii) to establish the physico-
chemical properties that inuence the rate constant of the
short Fe–O distance that has not dramatically changed from
chemical reaction.
what it was in the iron(IV)–oxo porphyrin cation radical state. As
expected the metal is considerably displaced from the plane
through the four nitrogen atoms of the porphyrin ring by as
Let us rst start with a comparison of the experimental and
computational reaction rates. As FT-ICR MS experiments are
being performed at very low pressures, these experimental
conditions are close to Ideal Gas conditions with very few
˚
much as 0.268–0.299 A. These transition states bear
Fig. 5 UB3LYP/BS1 optimized geometries of epoxidation transition states with bond lengths in angstroms.
1524 | Chem. Sci., 2015, 6, 1516–1529
This journal is © The Royal Society of Chemistry 2015

View Article Online
Edge Article
Chemical Science
molecular collisions per second. In the kinetic study of ion- experimental enthalpies of activation well and the linearity and
molecule reactions in the gas phase one needs to consider that reproducibility of the calculations is well within the typical error
thermal equilibration of the reacting system with the environ- reported for DFT calculations using this method of about 5 kcal
ment is in general not granted. On the contrary, when small molꢀ1 39 There is, however, a large systematic error as well as a
.
species react at low pressures, the absence of thermalizing relatively large standard deviation that require further studies.
collisions leads to non-equilibrium energy distributions. In the Note that the experimental data in Fig. 6 refers to free energies
absence of solvation, the double-well potential model rst of activation, whereas the computational results are enthalpy
proposed by Brauman in 1977 to account for the kinetic changes instead. The systematic error between experiment and
behaviour of displacement reactions by anionic nucleophiles theory contains entropic corrections to the energy.
predicts that the energy of the intermediate and transition state
Subsequently, we investigated the origin of the rate constant,
must lie below the energy of the combined reactants.38 Because and in particular, the physical and chemical properties of the
at low pressure the intermediates cannot be stabilized by substrate and oxidant that determine the reaction mechanism
unreactive collisions, determining the transition state energy is and the enthalpy of activation of an epoxidation reaction.
less straightforward than in solution. However, the kinetics Previous studies on heteroatom oxidation and double bond
results presently reported deal with a relatively large reactant epoxidation by P450 enzymes implicated a correlation between
ion that effectively establishes thermal equilibrium with the the natural logarithm of the rate constant with the ionization
environment through coupling with the background radiation energy of the substrate.12d,e,28,40 To nd out whether the data in
eld allowed by the several low frequency infrared modes of the Table 1 follow these trends as well, we plot RT ln kexp versus
iron(IV)–oxo macrocyclic ligand complex. This condition is experimentally known ionization energies,24 in Fig. 7. The set of
responsible, for example, for the consistency between the data shown in Table 1 and Fig. 7 gives a linear correlation
kinetics of NO ligand addition to iron(II/III) porphyrin complexes between the natural logarithm of the rate constant and the
and the equilibrium data independently established through ionization energy of the substrate with an R2 ¼ 0.96. Fig. 7b
equilibrium measurements.25 Because of these considerations, displays the correlation between the DFT calculated enthalpy of
the notion can be adopted that the presently investigated
systems are in prevailing thermal equilibrium with the envi-
ronment and reaction kinetics can be interpreted within the
framework of transition state theory. Consequently, reaction
rates represent bimolecular reactions and as such they should
compare to computationally determined reaction rates well.
The experimental rate constants (Table 1) were converted
into free energy units via ꢀRT ln kexp, with R the gas constant
and T the temperature, using transition state theory and plotted
against the calculated enthalpy of activation for the same
substrates, see Fig. 6. Although only a limited computational
study is reported here, when we calculate the deviation between
experiment and theory for each data point, we nd an average
difference between experiment and theory of 1.5 kcal molꢀ1
with a standard deviation of 3.4 kcal molꢀ1. As such, the DFT
methods used here reproduce the trends obtained from
Fig. 7 (a) Correlation between experimentally determined RT ln kexp
(for raw data, see Table 1) versus known ionization energies (IE). (b)
Correlation between calculated epoxidation activation enthalpy (in
kcal molꢀ1) and experimental ionization energy for the substrates in
Fig. 6 Correlation between experimental and computational barrier
heights.
Fig. 5.
This journal is © The Royal Society of Chemistry 2015
Chem. Sci., 2015, 6, 1516–1529 | 1525

View Article Online
Chemical Science
Edge Article
activation of the reaction of [FeIV(O)(Por+c)]+ with olens. In excited state in the reactant geometry (J*R), black line in Fig. 8.
agreement with the experimental trends given in part (a) of These two VB curves cross and lead to an avoided crossing and a
Fig. 7 also the computational trends link the natural logarithm transition state for the C–O bond formation with barrier DE‡.
of the rate constant to the ionization energy of the substrate. The barrier height is linearly proportional to the curve crossing
Clearly, the key physicochemical property that drives the reac- energy, which in its own right is a fraction of the excitation
tion mechanism and affects the rate constant of substrate energy (GH) from the reactant wave function to the product wave
epoxidation by iron(IV)–oxo porphyrin cation radical complexes function in the geometry of the reactants, i.e. for JR / J*R. The
is the ionization energy of the substrate.
difference in VB structures for JR and J*R thereby should give a
In order to explain the experimental and computational reection of the key electron transfer/migrations upon product
trends in the reaction mechanisms, we devised a valence bond formation. Moreover, based on the excitation energy, the factors
(VB) curve crossing diagram, which is schematically depicted in that determine the barrier height can be predicted.
Fig. 8. This diagram starts bottom le with the reactant
An analysis of the differences between the reactant and
conguration of [FeIV(O)(TPFPP+c)]+ in electronic conguration product wave functions in the reactant geometry reveals the
2
2
2
1
1
pxz pyz dx2ꢀy2 p*xz py*z a2u1. The p and p* electrons along the following information: First of all, a comparison of the VB
FeO bond are identied with dots in the VB diagram and due to structures of JR and J*R shows that the electrons in the p-bond
2
occupation of pxz p*xz1 there are three dots on the le-hand-side of the olen are singlet paired in the ground state and triplet
of the Fe–O bond. In addition, there are three electrons in the coupled in the excited state, hence the excitation energy GH
pyz and p*yz orbitals, which are identied with the other three includes the p–p* electron excitation in the substrate, Eex,Sub
.
dots on the right-hand-side of the Fe–O bond. Furthermore, the Generally, the rst ionization potential of an olen corresponds
oxidant has a radical on the porphyrin ring for single occupa- to the removal of an electron from a p-orbital, and, hence, is
tion of the a2u molecular orbital. The substrate double bond is proportional to the p–p* excitation energy. Indeed, our exper-
also highlighted with four electrons spread out over the inter- imentally and computationally determined barrier heights
action. Upon approach of the substrate on the iron(IV)–oxo correlate linearly with the ionization energy of the olen, and
species a radical intermediate is formed that has a single bond therefore support the VB model.
between the oxygen and carbon atoms and a doubly occupied
a2u orbital.
One of the electrons originating from the p-bond of the
olen forms a bond with the p*xz electron along the FeO bond, to
In VB theory the electronic conguration in the reactant create the C–O bonding pair of electrons. This means that the
complex (JR) connects to an excited state in the product pxz/p*xz pair of orbitals during the reaction splits back into
geometry (J*P) as shown with the blue line in Fig. 8. At the same individual atomic orbitals namely 3dxz(Fe) and 2px(O). The
time the product electronic conguration (JP) connects to an 2px(O) electron pairs with the electron from the substrate, while
one of the electrons of the 3dxz(Fe) orbital is transferred into the
a2u orbital through internal excitation/rehybridization of the
oxidant, Eex,ox. The promotion gap, GH, therefore, will be
proportional to the p–p* excitation in the substrate and the 3dxz
to a2u electron transfer in the oxidant: GH ¼ Eex,Sub + Eex,ox
.
Obviously, since the ionization energy represents the energy to
remove an electron from a p-type orbital of an olen, this will
imply a linear correlation between the rst excited state and the
ionization energy of the substrate.28 The VB diagram, therefore,
conrms a linear correlation between the ionization energy of
the substrate and the C–O bond formation enthalpy of activa-
tion as shown above.
Although, the oxygen atom transfer reaction between
[FeIV(O)(TPFPP+c)]+ and an olen could lead to either epoxide or
hydroxylated products, unfortunately the FT-ICR MS experi-
ments cannot distinguish the two. Thus, several substrates in
Scheme 2 and Table 1 contain aliphatic groups that in a reaction
with an iron(IV)–oxo group can be converted into an alcohol. A
correlation between the rate constant of oxygen atom transfer
and the ionization energy, Fig. 7a, of the olen provides indirect
experimental evidence that all reactions lead to epoxidation
products. In fact, hydrogen atom abstraction reactions should
not correlate with the ionization potential of the substrate, but
were shown to be proportional to the strength of the C–H bond
of the substrate that is formed.12e,40 To test that the rate
constants do not correlate with the bond dissociation energy
(BDECH) of the C–H bond of the substrate that is broken, we plot
Fig. 8 VB curve crossing diagram for the C–O bond formation step in
olefin epoxidation (R2C]CH2) by [FeIV(O)(TPFPP+c)]+. Valence elec-
trons are identified with a dot and lines (curved and straight) in the VB
structures represent bonds.
1526 | Chem. Sci., 2015, 6, 1516–1529
This journal is © The Royal Society of Chemistry 2015

View Article Online
Edge Article
Chemical Science
in Fig. 9 calculated BDECH and barrier heights of selected epoxidation by iron(IV)–oxo porphyrin cation radical
olens, namely propene, Z-2-butene, E-2-butene, cyclohexene complexes in the gas phase. We present a novel method to
and 1,3-cyclohexadiene. As can be seen from Fig. 9, no corre- synthesize [FeIV(O)(TPFPP+c)]+ in the gas phase at low pres-
lation between BDECH and barrier height exists, and, therefore, sure. Furthermore, we report a large set of experimentally
hydrogen atom abstraction is not the rate determining step in derived rate constants and product distributions. All olens
the reaction mechanism. Further evidence that hydrogen atom undergo oxygen atom transfer, whereas compounds with low
abstraction reactions can be ruled out here comes from kinetic ionization energy also give a certain degree of hydride transfer
isotope effect (KIE) studies. We measured the rate constant of and charge transfer reactions. Our experimentally determined
oxygen atom transfer with cyclohexene and cyclohexene-d10 and reaction rates correlate linearly with the ionization potential of
determined a KIE ¼ kH/kD ꢄ1 (Table 1). Consequently, the the substrate and show that the electron transfer from substrate
oxygen atom transfer is unlikely to proceed with an initial to oxidant is rate determining. A thorough computational
hydrogen atom abstraction and double bond epoxidation will survey has conrmed the suggested mechanism and provides a
be the dominant pathway.
rationale for the observed trend in the rate constants. Moreover,
The experimental trends, therefore, provide the rst indirect the work highlights a regioselective epoxidation over hydroxyl-
experimental evidence that in the gas-phase the regioselectivity ation in the gas phase.
of double bond epoxidation versus aliphatic hydroxylation will
be in favour of the epoxidation pathway. This implies that in
enzymatic systems, such as the cytochromes P450, the shape
Acknowledgements
and size of the substrate binding pocket will inuence the MAS thanks the Petroleum Technology Development Fund for a
regioselectivity of hydroxylation over epoxidation and can studentship. The National Service of Computational Chemistry
change the natural preference away from epoxidation.
Soware (NSCCS) is thanked for providing valuable cpu time to
Finally, the calculations presented in this work obviously SdV. DK acknowledges the Department of Science and Tech-
refer to gas-phase results and hence correlate well with gas- nology (New Delhi) for a Ramanujan Fellowship. MEC and SF
`
phase mass spectrometric data. In order to further establish thank the Universita degli Studi di Roma La Sapienza for
that the work can be extrapolated to solution phase, we did a support.
series of single point calculations using the polarized
continuum model with a dielectric constant of 3 ¼ 4.7 to mimic
Notes and references
a
solution. The obtained correlation between solvent
corrected free energies of activation of epoxidation reactions by
[FeIV(O)(Por+c)]+ is plotted against the solvent corrected ioniza-
tion energy of all substrates in Fig. S1, ESI.† Even in solvent, the
linear trend in the correlation between free energy of activation
and ionization energy is retained, therefore, we expect to be able
to extrapolate our results to the solution phase as well.
1 (a) M. Sono, M. P. Roach, E. D. Coulter and J. H. Dawson,
Chem. Rev., 1996, 96, 2841; (b) J. T. Groves, Proc. Natl. Acad.
Sci. U. S. A., 2003, 100, 3569; (c) Cytochrome P450: Structure,
Mechanism and Biochemistry, ed. P. R. Ortiz de Montellano,
Kluwer Academic/Plenum Publishers, New York, 3rd edn,
2004; (d) I. G. Denisov, T. M. Makris, S. G. Sligar and
I. Schlichting, Chem. Rev., 2005, 105, 2253; (e) Handbook of
Porphyrin Science, ed. K. M. Kadish, K. M. Smith and R.
Guilard, World Scientic, New Jersey, 2010; (f) P. R. Ortiz
de Montellano, Chem. Rev., 2010, 110, 932; (g) E. O'Reilly,
V. Koehler, S. Flitsch and N. Turner, Chem. Commun., 2011,
47, 2490; (h) G. Grogan, Curr. Opin. Chem. Biol., 2011, 15, 241.
2 (a) J. T. Groves, Models and Mechanisms of Cytochrome
P450 Action, in Cytochrome P450: Structure, Mechanism and
Biochemistry, ed. P. R. Ortiz de Montellano, Kluwer
Academic/Plenum Publishers, New York, 3rd edn, 2005, ch.
1, pp. 1–44; (b) Y. Watanabe, H. Nakajima and T. Ueno,
Acc. Chem. Res., 2007, 40, 554.
Conclusions
In this work we report a comprehensive combined mass
spectrometric and computational study on substrate
3 D. R. Nelson, Methods Mol. Biol., 2006, 320, 1.
4 (a) T. L. Poulos, B. C. Finzel, I. C. Gunsalus, G. C. Wagner and
J. Kraut, J. Biol. Chem., 1985, 260, 6122; (b) T. L. Poulos,
B. C. Finzel and A. J. Howard, Biochemistry, 1986, 25, 5314.
5 J. B. Johnston, P. M. Kells, L. M. Podust and P. R. Ortiz de
Montellano, Proc. Natl. Acad. Sci. U. S. A., 2009, 106, 20687.
6 (a) B. Meunier, S. P. de Visser and S. Shaik, Chem. Rev., 2004,
104, 3947; (b) Iron-containing enzymes: Versatile catalysts of
hydroxylation reaction in nature, ed. S. P. de Visser and D.
Kumar, RSC Publishing, Cambridge, UK, 2011.
Fig.
9 Correlation between calculated epoxidation activation
enthalpy (in kcal molꢀ1) and BDECH for the substrates.
7 J. Rittle and M. T. Green, Science, 2010, 330, 9337.
This journal is © The Royal Society of Chemistry 2015
Chem. Sci., 2015, 6, 1516–1529 | 1527

View Article Online
Chemical Science
Edge Article
8 (a) K. Auclair, Z. Hu, D. M. Little, P. R. Ortiz de Montellano 18 (a) R. Lati, M. A. Sainna, E. V. Rybak-Akimova and S. P. de
and J. T. Groves, J. Am. Chem. Soc., 2002, 124, 6020; (b)
Visser, Chem.–Eur. J., 2013, 19, 4058; (b) M. G. Quesne,
R. Lati, L. E. Gonzalez-Ovalle, D. Kumar and S. P. de
Visser, Chem.–Eur. J., 2014, 20, 435.
´
P. Late, F. Andre, D. C. Zeldin, P. M. Dansette and
D. Mansuy, Biochemistry, 2007, 46, 10237; (c) T. S. Dowers,
D. A. Rock, D. A. Rock and J. P. Jones, J. Am. Chem. Soc., 19 (a) A. D. Becke, J. Chem. Phys., 1993, 98, 5648; (b) C. Lee,
2004, 126, 8868; (d) H. L. R. Cooper and J. T. Groves, Arch.
Biochem. Biophys., 2011, 507, 111; (e) K. M. Roberts and
J. P. Jones, Chem.–Eur. J., 2010, 16, 8096.
9 See, e.g., (a) J. T. Groves and R. S. Myers, J. Am. Chem. Soc.,
1983, 105, 5791; (b) J. T. Groves and Y. Watanabe, J. Am.
W. Yang and R. G. Parr, Phys. Rev. B: Condens. Matter
Mater. Phys., 1988, 37, 785.
20 (a) A. K. Vardhaman, C. V. Sastri, D. Kumar and S. P. de
Visser, Chem. Commun., 2011, 47, 11044; (b)
A. K. Vardhaman, P. Barman, S. Kumar, C. V. Sastri,
D. Kumar and S. P. de Visser, Angew. Chem., Int. Ed., 2013,
52, 12288; (c) S. Kumar, A. S. Faponle, P. Barman,
A. K. Vardhaman, C. V. Sastri, D. Kumar and S. P. de
Visser, J. Am. Chem. Soc., 2014, 136, 17102.
´
Chem. Soc., 1986, 108, 507; (c) D. Ostovic and T. C. Bruice,
J. Am. Chem. Soc., 1989, 111, 6511; (d) J. T. Groves, Z. Gross
and M. K. Stern, Inorg. Chem., 1994, 33, 5065; (e)
N. Stephenson and A. T. Bell, Inorg. Chem., 2006, 45, 2758;
(f) N. Hessenauer-Ilicheva, A. Franke, D. Meyer, 21 P. J. Hay and W. R. Wadt, J. Chem. Phys., 1985, 82, 270.
W.-D. Woggon and R. van Eldik, J. Am. Chem. Soc., 2007, 22 S. Grimme, J. Antony, S. Ehrlich and H. Krieg, J. Chem. Phys.,
129, 12473; (g) R. Mas-Balleste and L. Que Jr, J. Am. Chem.
2010, 132, 154104.
Soc., 2007, 129, 15964; (h) P. Comba and G. Rajaraman, 23 S. P. de Visser, J. Am. Chem. Soc., 2010, 132, 1087.
Inorg. Chem., 2008, 47, 78; (i) P. C. A. Bruijnincx, 24 S. G. Lias, J. E. Bartmess, J. F. Liebman, J. L. Holmes,
I. L. C. Buurmans, S. Gosiewska, M. A. H. Moelands,
M. Lutz, A. L. Spek, G. van Koten and R. J. M. Klein
Gebbink, Chem.–Eur. J., 2008, 14, 1228; (j) J. F. Hull,
E. L. O. Sauer, C. D. Incarvito, J. W. Faller, G. W. Brudvig
and R. H. Crabtree, Inorg. Chem., 2009, 48, 488; (k)
A. McGown, W. D. Kerber, H. Fujii and D. P. Goldberg, J.
R. D. Levin and W. G. Mallard, in Ion Energetics Data in
NIST Chemistry WebBook, ed. P. J. Linstrom and W. G.
Mallard, National Institute of Standards and Technology,
Gaithersburg, MD, 20899, http://webbook.nist.gov, NIST
Standard Reference Database Number 69, retrieved in
September 2014.
Am. Chem. Soc., 2009, 131, 8040; (l) A. Company, Y. Feng, 25 F. Angelelli, B. Chiavarino, M. E. Crestoni and S. Fornarini, J.
¨
M. Guell, X. Ribas, J. M. Luis, L. Que Jr and M. Costas,
Am. Soc. Mass Spectrom., 2005, 16, 589.
Chem.–Eur. J., 2009, 15, 3359; (m) A. Franke, M. Wolak and 26 S. P. de Visser, Inorg. Chem., 2006, 45, 9551.
R. van Eldik, Chem.–Eur. J., 2009, 15, 10182; (n) 27 (a) R. E. White, J. T. Groves and G. A. McClusky, Acta Biol.
N. Hessenauer-Ilicheva, A. Franke, M. Wolak, T. Higuchi
and R. van Eldik, Chem.–Eur. J., 2009, 15, 12447.
10 (a) Z. Gross and S. Nimri, Inorg. Chem., 1994, 33, 1731; (b)
Z. Gross, J. Biol. Inorg. Chem., 1996, 1, 368; (c)
K. Czarnecki, S. Nimri, Z. Gross, L. M. Proniewicz and
J. R. Kincaid, J. Am. Chem. Soc., 1996, 118, 2929.
Med. Ger., 1979, 38, 475; (b) R. T. Ruettinger and
A. J. Fulco, J. Biol. Chem., 1981, 256, 5728; (c) J. T. Groves
and D. V. Subramanian, J. Am. Chem. Soc., 1984, 106, 2177;
(d) J. T. Groves, G. E. Avaria-Neisser, K. M. Fish, M. Imachi
and R. L. Kuczkowski, J. Am. Chem. Soc., 1986, 108, 3837;
(e) J. F. Bartoli, O. Brigaud, P. Battioni and D. Mansuy, J.
Chem. Soc., Chem. Commun., 1991, 440; (f) T. Ohno,
N. Suzuki, T. Dokoh, Y. Urano, K. Kikuchi, M. Hirobe,
T. Higuchi and T. Nagano, J. Inorg. Biochem., 2000, 82, 123;
(g) S. Yoshioka, S. Takahashi, K. Ishimori and
I. Morishima, J. Inorg. Biochem., 2000, 81, 141.
11 (a) A. G. Marshall, C. L. Hendrickson and G. S. Jackson, Mass
´
¨
Spectrom. Rev., 1998, 17, 1; (b) J. Roithova and D. Schroder,
Chem. Rev., 2010, 110, 1170; (c) N. M. M. Nibbering, Mass
Spectrom. Rev., 2006, 25, 962.
12 (a) M. E. Crestoni and S. Fornarini, Inorg. Chem., 2005, 44,
5379; (b) M. E. Crestoni and S. Fornarini, Inorg. Chem., 28 S. P. de Visser, F. Ogliaro, P. K. Sharma and S. Shaik, J. Am.
2007, 46, 9018; (c) M. E. Crestoni, S. Fornarini, Chem. Soc., 2002, 124, 11809.
F. Lanucara, J. J. Warren and J. M. Mayer, J. Am. Chem. 29 (a) M. T. Green, J. Am. Chem. Soc., 1999, 121, 7939; (b)
Soc., 2010, 132, 4336; (d) M. E. Crestoni, S. Fornarini and
F. Lanucara, Chem.–Eur. J., 2009, 15, 7863; (e) F. Lanucara
and M. E. Crestoni, Chem.–Eur. J., 2011, 17, 12092.
13 H. Saltzman and J. G. Sharein, Organic Syntheses, Wiley,
New York, 1973, vol. 5, p. 658, Collect.
F. Ogliaro, S. P. de Visser, S. Cohen, J. Kaneti and S. Shaik,
ChemBioChem, 2001, 11, 848.
30 J. H. Dawson, R. H. Holm, J. R. Trudell, G. Barth,
R. E. Linder, E. Bunnenberg, C. Djerassi and S. C. Tang, J.
Am. Chem. Soc., 1976, 98, 3707.
14 (a) M. Meot-Ner, in Gas Phase Ion Chemistry, ed. M. T. 31 (a) S. P. de Visser, S. Shaik, P. K. Sharma, D. Kumar and
Bowers, Academic Press, New York, 1979, vol. 1; (b)
J. E. Bartmess and R. M. Georgiadis, Vacuum, 1983, 33, 149.
15 T. Su and W. J. Chesnavich, J. Chem. Phys., 1982, 76, 5183.
16 B. Chiavarino, R. Cipollini, M. E. Crestoni, S. Fornarini,
F. Lanucara and A. Lapi, J. Am. Chem. Soc., 2008, 130, 3208.
17 (a) Schrodinger, Jaguar, version 7.9, LLC, New York, NY, 2011;
(b) M. J. Frisch, Gaussian 09, revision D.01, Gaussian, Inc.,
Wallingford, CT, 2004.
W. Thiel, J. Am. Chem. Soc., 2003, 125, 15779; (b)
T. Kamachi and K. Yoshizawa, J. Am. Chem. Soc., 2003, 125,
4652; (c) C. M. Bathelt, J. Zurek, A. J. Mulholland and
J. N. Harvey, J. Am. Chem. Soc., 2005, 127, 12900; (d)
S. P. de Visser, J. Phys. Chem. A, 2005, 109, 11050; (e)
S. P. de Visser, L. Tahsini and W. Nam, Chem.–Eur. J.,
´
2009, 15, 5577; (f) R. Lonsdale, J. Olah, A. J. Mulholland
and J. N. Harvey, J. Am. Chem. Soc., 2011, 133, 15464; (g)
1528 | Chem. Sci., 2015, 6, 1516–1529
This journal is © The Royal Society of Chemistry 2015

View Article Online
Edge Article
H. Isobe, S. Yamanaka, M. Okumura, K. Yamaguchi and
J. Shimada, J. Phys. Chem. B, 2011, 115, 10730.
32 A. Ghosh, Acc. Chem. Res., 1998, 31, 189.
Chemical Science
Angew. Chem., Int. Ed., 2002, 41, 1947; (d) D. Kumar,
B. Karamzadeh, G. N. Sastry and S. P. de Visser, J. Am.
Chem. Soc., 2010, 132, 7656; (e) D. Kumar, S. P. de Visser
and S. Shaik, Chem.–Eur. J., 2005, 11, 2825.
33 (a) H. M. Neu, M. G. Quesne, T. Yang, K. A. Prokop-Prigge,
K. M. Lancaster, J. Donohoe, S. DeBeer, S. P. de Visser and 37 See, e.g., (a) A. Ghosh and P. R. Taylor, Curr. Opin. Chem.
D. P. Goldberg, Chem.–Eur. J., 2014, 20, 14584; (b) F. Lanucara,
B. Chiavarino, M. E. Crestoni, D. Scuderi, R. K. Sinha,
P. Maitre and S. Fornarini, Inorg. Chem., 2011, 50, 4445.
Biol., 2003, 7, 113; (b) M. Swart, A. R. Groenhof,
A. W. Ehlers and K. Lammertsma, J. Phys. Chem. A, 2004,
108, 5479; (c) S. P. de Visser, M. G. Quesne, B. Martin,
P. Comba and U. Ryde, Chem. Commun., 2014, 50, 262.
38 M. L. Chabinyc, S. L. Craig, C. K. Regan and J. I. Brauman,
Science, 1998, 279, 1882.
´
34 M. Radon, E. Broclawic and K. Pierloot, J. Chem. Theory
Comput., 2011, 7, 898.
35 (a) D. Kumar, L. Tahsini, S. P. de Visser, H. Y. Kang, S. J. Kim
and W. Nam, J. Phys. Chem. A, 2009, 113, 11713; (b) 39 (a) Y. Zhao and D. G. Truhlar, Acc. Chem. Res., 2008, 41, 157;
D. Kumar, R. Lati, S. Kumar, E. V. Rybak-Akimova,
M. A. Sainna and S. P. de Visser, Inorg. Chem., 2013, 52, 7968.
36 (a) S. P. de Visser, F. Ogliaro, N. Harris and S. Shaik, J. Am.
(b) T. Schwabe and S. Grimme, Acc. Chem. Res., 2008, 41, 569;
(c) T. Weymuth, E. P. A. Couzijn, P. Chen and M. Reiher, J.
Chem. Theory Comput., 2014, 10, 3092.
Chem. Soc., 2001, 123, 3037; (b) S. P. de Visser, F. Ogliaro 40 Y. Watanabe, T. Iyanagi and S. Oae, Tetrahedron Lett., 1980,
and S. Shaik, Angew. Chem., Int. Ed., 2001, 40, 2871; (c)
21, 3685.
S. P. de Visser, F. Ogliaro, P. K. Sharma and S. Shaik,
This journal is © The Royal Society of Chemistry 2015
Chem. Sci., 2015, 6, 1516–1529 | 1529