Resonant inelastic X-ray scattering on ferrous and ferric bis-imidazole porphyrin and cytochrome c: Nature and role of the axial methionine-fe bond

Article abstract of DOI:10.1021/ja5100367

Axial Cu-S(Met) bonds in electron transfer (ET) active sites are generally found to lower their reduction potentials. An axial S(Met) bond is also present in cytochrome c (cyt c) and is generally thought to increase the reduction potential. The highly covalent nature of the porphyrin environment in heme proteins precludes using many spectroscopic approaches to directly study the Fe site to experimentally quantify this bond. Alternatively, L-edge X-ray absorption spectroscopy (XAS) enables one to directly focus on the 3d-orbitals in a highly covalent environment and has previously been successfully applied to porphyrin model complexes. However, this technique cannot be extended to metalloproteins in solution. Here, we use metal K-edge XAS to obtain L-edge like data through 1s2p resonance inelastic X-ray scattering (RIXS). It has been applied here to a bis-imidazole porphyrin model complex and cyt c. The RIXS data on the model complex are directly correlated to L-edge XAS data to develop the complementary nature of these two spectroscopic methods. Comparison between the bis-imidazole model complex and cyt c in ferrous and ferric oxidation states show quantitative differences that reflect differences in axial ligand covalency. The data reveal an increased covalency for the S(Met) relative to N(His) axial ligand and a higher degree of covalency for the ferric states relative to the ferrous states. These results are reproduced by DFT calculations, which are used to evaluate the thermodynamics of the Fe-S(Met) bond and its dependence on redox state. These results provide insight into a number of previous chemical and physical results on cyt c.

Full text of DOI:10.1021/ja5100367

Article
pubs.acs.org/JACS
Resonant Inelastic X‑ray Scattering on Ferrous and Ferric
Bis-imidazole Porphyrin and Cytochrome c: Nature and Role
of the Axial Methionine−Fe Bond
Thomas Kroll,†,‡ Ryan G. Hadt,† Samuel A. Wilson,† Marcus Lundberg,†,§ James J. Yan,†
Tsu-Chien Weng,⊥ Dimosthenis Sokaras,⊥ Roberto Alonso-Mori,‡ Diego Casa,∥ Mary H. Upton,∥
Britt Hedman,*,⊥ Keith O. Hodgson,*,†,⊥ and Edward I. Solomon*,†,⊥
†Department of Chemistry, Stanford University, Stanford, California 94305, United States
‡Linac Coherent Light Source, SLAC National Accelerator Laboratory, Stanford University, Menlo Park, California 94025, United States
§Department of Chemistry  Ångstrom, Uppsala University, SE-751 20 Uppsala, Sweden
̈
⊥Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, Stanford University, Menlo Park,
California 94025, United States
∥Advanced Photon Source, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, Illinois 60439, United States
S
* Supporting Information
ABSTRACT: Axial Cu−S(Met) bonds in electron transfer (ET) active sites are generally found to lower their reduction
potentials. An axial S(Met) bond is also present in cytochrome c (cyt c) and is generally thought to increase the reduction
potential. The highly covalent nature of the porphyrin environment in heme proteins precludes using many spectroscopic
approaches to directly study the Fe site to experimentally quantify this bond. Alternatively, L-edge X-ray absorption spectroscopy
(XAS) enables one to directly focus on the 3d-orbitals in a highly covalent environment and has previously been successfully
applied to porphyrin model complexes. However, this technique cannot be extended to metalloproteins in solution. Here, we use
metal K-edge XAS to obtain L-edge like data through 1s2p resonance inelastic X-ray scattering (RIXS). It has been applied here
to a bis-imidazole porphyrin model complex and cyt c. The RIXS data on the model complex are directly correlated to L-edge
XAS data to develop the complementary nature of these two spectroscopic methods. Comparison between the bis-imidazole
model complex and cyt c in ferrous and ferric oxidation states show quantitative differences that reflect differences in axial ligand
covalency. The data reveal an increased covalency for the S(Met) relative to N(His) axial ligand and a higher degree of covalency
for the ferric states relative to the ferrous states. These results are reproduced by DFT calculations, which are used to evaluate the
thermodynamics of the Fe−S(Met) bond and its dependence on redox state. These results provide insight into a number of
previous chemical and physical results on cyt c.
understood:2−4 the presence of the axial S(Met) ligand lowers E0
INTRODUCTION
■
due to the stronger stabilization energy (i.e., larger axial ligand
A common structural motif across electron transfer (ET)
metalloprotein active sites is the presence of an axial metal-
bond strength) in the oxidized than in the reduced state,5 while
not greatly affecting the reorganization energy.2−4
thioether bond, where this ligand is provided by an endogenous
methionine residue.1 These are found in the mononuclear type
one copper (T1 Cu), binuclear CuA, and cytochrome (cyt c
and cyt b) containing ET metalloproteins. In T1 Cu and CuA
active sites, the function of the Cu−S(Met) bond is now well
Received: October 7, 2014
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Interestingly, in contrast to T1 Cu and CuA, the Fe−S(Met)
bond in cyt c is generally thought to be stronger in the reduced
(ferrous) rather than the oxidized (ferric) state. This
consideration for cyt c is based on the decrease in E0 when the
Fe−S(Met) bond is lost,6,7 ligand competition8,9 and bind-
ing,10−12 protein folding studies,10,13−15 and the shorter Fe(II)-
S(Met) bond (2.29 Å) relative to Fe(III)-S(Met) (2.33 Å) from
extended X-ray absorption fine structure (EXAFS).16,17 Loss or
rupture of the Fe−S(Met) bond in cyt c is also functionally
relevant. Besides carrying out ET, cyt c is also involved in
programmed cell death or apoptosis.18 During this process, cyt c
switches function from an ET protein to a peroxidase enzyme.
This functional switch is correlated to the loss of the axial
Fe−S(Met) bond and the formation of an open coordination
sphere for the activation of H2O2 in the peroxidation of
cardiolipin.19−21 In addition, the Fe(II)-S(Met) bond can be
photodissociated, while the Fe(III)-S(Met) bond cannot.22−25
Clearly, the difference in metal−ligand bonding, and thus bond
strengths for the Fe(II) and Fe(III) states, is directly related to
both functions of cyt c: ET and lipid peroxidation.
by applying Kα resonant inelastic X-ray scattering (1s2p
RIXS).31−36Here, a 1s electron is excited into an unoccupied
3d orbital (1s → 3d) via a quadrupole transition to a 1s12p63dn+1
intermediate state and the subsequent electric dipole allowed
decay of a 2p electron into the 1s hole (2p → 1s) to the
1s22p53dn+1 final state, is detected by its photon emission. Thus,
L-edge absorption and 1s2p RIXS reach the same final state
configurations34,37 and allow a direct comparison, but with
complementary selection rules.38 Varying the incident X-ray
energies and collecting the respective emission spectra, allows
one to generate the RIXS plane. “L-edge-like” spectra are
obtained as vertical cuts through the RIXS plane (i.e., at fixed
K-edge excitation energies), providing the energies of the same
final state configurations that could be reached through direct
L-edge XAS, but via an intermediate state.
It has been recently shown that the valence bond configuration
interaction (VBCI) model with the same parameters reproduces
both L-edge XAS and RIXS results, but with selection rules38
involving the intermediate states in RIXS, as described by the
Kramers−Heisenberg scattering equation.31,32
In order to understand these differences in bonding, it is
necessary to employ a spectroscopic method that directly
probes the electronic structure of the Fe site; however, the highly
covalent porphyrin obscures the optical spectral features
associated with the Fe−S(Met) bond. S K-edge X-ray absorption
spectroscopy (XAS) has proven to be a powerful spectroscopic
tool to quantitate the covalency of thiolate-metal bonds in
proteins and models through the intensity of the pre-edge feature
(i.e., the 1s → 3d transition(s)).26 However, the higher Zeff of the
thioether sulfur relative to a thiolate increases the energy of the
pre-edge transition such that it is located within the envelope of
the edge transitions (i.e., the 1s → valence np).
The direct way to investigate the electronic structure of a heme
protein, where the Fe is in the highly covalent porphyrin environ-
ment, would be to probe the unoccupied valence orbitals of the
central Fe ion by Fe L-edge XAS. This is a powerful method to
extract the differential orbital covalency (DOC, differences in the
covalency of the different symmetry sets of d orbitals)27−30 that
is particularly important for heme complexes and enzyme
intermediates. However, the Fe L-edge excitation energy lies in
the soft X-ray energy region (≈710 eV), which requires ultrahigh
vacuum conditions and generally involves the detection of
electrons within 10 Å of the surface of the sample due to the low
electron escape depth. The escape depth can be enlarged to a few
micrometers by detecting the fluorescence decay. However, this
can suffer from significant self-absorption effects, mainly for the
L3-edge. These restrictions make it challenging to apply high-
resolution metal L-edge XAS to metalloproteins and enzymes.
Metal K-edge XAS does not have these restrictions due to
the high excitation energy in the hard X-ray region (the Fe
K-pre-edge is at ≈7110 eV). However, due to the short lifetime
of the 1s core-hole, the broadening of the 1s → 3d transitions is
significantly larger than for the 2p → 3d transitions. Since the
1s → 3d transitions are quadrupole allowed, they often gain
intensity from small amounts of 4p mixing into the d orbitals
in noncentrosymmetric ligand environments, while 2p → 3d
transitions are electric dipole allowed and their intensities are a
direct probe of the metal d-character in valence orbitals involved
in bonding. Thus, by going from soft to hard X-rays, one wins on
experimental conditions, but loses resolution and insight into
bonding.
⟨f|H1s2p|i⟩⟨i|H1s3d|g⟩
F(Ω, ω) ∝
∑ ∑
iΓ
Eg − Ef + Ω −
f
i
2
×δ(Eg − Ef + Ω + ω)
(1)
where Ω and ω are the incident and emission energy,
respectively, with Ω − ω being the energy transfer, that gives
“L-edge-like” spectra.
To investigate the electronic structure of Fe in a covalent
porphyrin environment, and in particular the Fe−S(Met)
bonding properties of ferrous and ferric cyt c, 1s2p RIXS was
applied to this protein, which cannot be reasonably studied by
L-edge XAS. Parallel RIXS data were collected on tetraphenyl
porphyrin (TPP) complexes with bisimidazole axial ligands
(FeTPP(ImH)2. These complexes have already been studied by
L-edge XAS.29 The two main differences in the structures of cyt
c and FeTPP(ImH)2, include replacement of one axial imidazole
ligand with a thioether (Met-80), and the porphyrin ring is
cross-linked to two cysteine derived thioether bonds (cys-14 and
cys-17) (see Figure 1). Here, RIXS data on FeTPP(ImH)2 are
Figure 1. Molecular structures of (a) FeTPP(ImH)2, and (b) its active
side counterpart in cyt c.
correlated to the L-edge XAS data and to 1s2p RIXS data on
ferrous and ferric cyt c, and analyzed using the VBCI multiplet
model. These results are then used to correlate to DFT cal-
culations on large models to distinguish specific contributions to
bonding. Truncated models, which reasonably represent the
large models (see Figure S1 in Supporting Information), are then
used to explore the thermodynamic differences between the
The desired combination of accessible experimental con-
ditions, higher resolution, and d-orbital covalency can be realized
B
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Fe−S(Met) and Fe−N(His) bond in the ferrous and ferric states.
The results of these studies are then related to past experimental
results on thioether-Fe bonding.
EXPERIMENTAL AND COMPUTATIONAL METHODS
■
Sample Preparation. Ferrous and ferric FeTPP(ImH)2 and cyt
c were prepared in-house following standard procedures. Details about
the sample preparation are given in the Supporting Information.
Experimental Methods. Three experimental methods have been
employed: soft X-ray L-edge XAS, as well as hard X-ray K-edge XAS and
1s2p RIXS. These experiments have been performed at the Stanford
Synchrotron Radiation Lightsource (SSRL) at beamlines 10−1(L-edge
XAS), 7−3 (K-edge) and 6−2 (1s2pRIXS),39 as well as the Advanced
Photon Source beamline 9-ID.40,41 Experimental details on all methods
are given in the Supporting Information. All data sets (L-edge, K-edge,
1s2p RIXS) have been measured at minimum twice to ensure
reproducible and reliable data.
The pre-edge in a 1s2p RIXS experiment includes the tail of the
intense 1s → 4p excitations at around 15 eV higher energy. This leads to
an increase of the pre-edge background and further complicates a direct
comparison of L-edge data with constant incident energy (CIE, i.e.,
vertical) cuts through the RIXS plane. In order to estimate the RIXS
intensity from the tail of the rising-edge, the data were carefully fit to
extract the true pre-edge region as described in the Supporting
Information and in the literature.37
Computational Methods. To simulate all X-ray spectra, charge
transfer multiplet calculations, that properly take the core-hole into
account, were performed.42−45 Covalent mixing of ligand character is
modeled using a valence bond configuration interaction (VBCI)
approach, using a three configuration model including a ligand-to-
metal charge transfer (LMCT) configuration dn+1L and a metal-to-
ligand charge transfer (MLCT) configuration dn−1L− mixed into the dn
configuration in the ground state. (L = ligand with an e− hole, L− = ligand
with an additional electron) The corresponding charge transfer energies
are Δ (LMCT) and Δ* (MLCT). All calculations have been performed
in D4h effective symmetry for all four complexes. In the ferrous case, the
x2 − y2/z2 energy splitting and covalencies are properly reflected in this
approximation. In the ferrous case, the x2 − y2/z2 energy splitting is
reflected properly in D4h symmetry. In the ferric case, there is one 3d?
hole, and thus the quantification only reflects the covalency of this hole.
The covalency value for each of the symmetry blocks was generated via
the projection method27 leading to the differential orbital covalency
(DOC).
Figure 2. Fe(II)TPP(ImH)2: (a) Full 1s2p RIXS plane, (b) rising-edge
subtracted 1s2p RIXS plane, (c) comparison of the CEE cut through
the uncorrected plane (red line in (a)) with the direct K-edge,
(d) comparison of the CIE cut along the red line in (b) with the direct
L-edge experimental data.
Ground state DFT calculations and geometry optimizations were
performed using Gaussian 0946 for a range of functionals and basis sets.
Unoccupied metal d-character has been determined by subtracting the
sum of their contribution to all occupied orbitals from 100%. The ligand
donor character has been obtained from the occupied fragment orbital
character summed over the unoccupied orbitals. Further details are
given in the Supporting Information.
Figure 2(c) compares the K-edge XAS of the Fe(II)TPP-
(ImH)2 model complex with the constant emission energy
(CEE) cut through the RIXS plane as indicated by the diagonal
red line in Figure 2(a). Both spectra show the same structure that
is better resolved in the CEE spectrum (this is often also referred
to as HERFD). The peak at 7112.2 eV involves excitation to the
1s13d6t2g3de1g final state, while the second weak peak at 7115.3 eV
likely originates from π back-bonding. The existence of a small π
back-bonding contribution (10%) in the ground state has already
beenobserved inL-edge XAS and is present in DFT calculations.29
The x2 − y2 and z2 orbitals are separated in energy due to the
difference in axial and equatorial bonding, associated with effec-
tive D4h symmetry. Since the xy and xz/yz orbitals are occupied,
the K-pre-edge is dominated by excitations into these empty
x2 − y2 and z2 orbitals. To estimate the maximum D4h final state
splitting, the Fe(II)TPP(ImH)2 result was compared to the 1s2p
RIXS of Fe(II)-tacn, which has Oh local symmetry and thus only
one possible K-pre-edge peak (see Figure S2 in the Supporting
Information).47 This comparison limits the possible eg peak
splitting to ≤0.4 eV.
RESULTS AND ANALYSIS
■
Ferrous Spectroscopy. Ferrous TPP-bisIm. Figure 2
presents a collection of data sets comparing the 1s2p RIXS,
L-edge and K-edge XAS data. Figure 2(a) shows the full 1s2p
RIXS plane, while Figure 2(b) gives the result with the rising
edge subtracted as described in the Supporting Information. The
features at an incident energy of ≈7112 eV and energy transfers
of ≈706.5 and ≈720 eV reflect excitations into the K-pre-edge.
The L3- and L2-edge-like features on the energy transfer axis are
split by the spin−orbit coupling within the 2p core-hole in the
final state. The intense feature starting at ≈7115 eV originates
from the tail of the strong dipole allowed 1s−4p transition
at ≈7130 eV. Note that possible shakeup features due to
σ+π-donation and π-back-bonding could be present at the onset
of the main-edge tail that have been neglected in the background
subtracted RIXS in Figure 2(b).
C
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
While no spectroscopic difference is observed between the
K-edge XAS and CEE RIXS, this is not the case in comparing the
L-edge XAS data with a CIE cut at hν = 7112.2 eV through the
RIXS plane (red vertical line in Figure 2(b)). Figure 2(d)
presents the Fe(II)TPP(ImH)2 L-edge XAS data together with
the CIE cut through the main RIXS peak.
The CIE cut shows additional intensity at lower energies for
both the L3- and L2-like curves (hν = 707.3 and 719.6 eV) and a re-
duced relative intensity at the L-edge main peak at hν = 708.2 eV.
Thehigh energy L3-edge shoulder inthe CIE cut at710.2 eV is also
less pronounced compared to the corresponding L-edge feature.
Similar observations hold true for the L2-edge energy region,
together with the observation of a smaller L3−/L2− peak intensity
ratio in comparing the CIE RIXS cut to the direct L-edge XAS for
Fe(II)TPP(ImH)2.
In our earlier 1s2p RIXS experiments on Oh samples,
additional intensity in the CIE RIXS cuts relative to L-edge
XAS was found to originate from electric dipole forbidden excita-
tions38 that are allowed in 1s2p RIXS (see eq 1 and Figure 3(a)).
Figure 3. Possible excitation pathways for an L-edge and 1s2p RIXS
experiment of a ferrous low-spin system for (a) Oh and (b) D4h
symmetry. The pathways through B2g and Eg intermediate states in
(b) have been omitted for clarity.
In the Fe(II)TPP(ImH)2 model complex, the Fe local symmetry
is approximately D4h with an A1g ground state. In D4h symmetry,
the L-edge electric dipole transition operator has a2u and eu sym-
metry so that only A2u and Eu final states are allowed(Figure 3(b)).
In the 1s2p RIXS experiment, one excites via a 1s → 3d quadrupole
transition, having a1g and b1g character, into intermediate states
with A1g (Fe dz ) and B1g (Fe dx −y ) symmetry, respectively.
The 2p → 1s decay from these final states again has electric dipole
character, thus a2u and eu symmetry, which allows 1s2pRIXSinten-
sity to final states with B2u, A2u and Eu symmetry (Figure 3(b)).
Final states with Eu symmetry can be reached both from A1g and
B1g intermediate states resulting in different final state energies.
Thus, the additional intensity found in the 1s2p RIXS cuts
throughout the main intensity peaks originates from both final
states that cannot be reached in L-edge XAS and also from the
splitting of the x2 − y2/z2 orbitals in the effective D4h symmetry of
the porphyrin complex.
Ferrous cyt c. Figure 4 presents the 1s2p RIXS plane of ferrous
cyt c. Figure 4(a) shows the full RIXS plane as measured, while
in (b) the rising edge has been subtracted. The pre-edge peak
structure contains mainly one broad intensity region at the
L3- and L2-like energies, with an intense tail to higher energies.
This tail however is weaker in intensity compared to the ferrous
model RIXS plane (Figure 2(b)). In (c) we compare the CEE
cuts of ferrous cyt c with that of the Fe(II)TPP(ImH)2 complex.
Both have very similar pre-edge intensities and peak widths, as
well as additional intensity at ≈7115.3 eV, likely due to π back-
bonding. However, ferrous cyt c shows a steeper rising edge,
which overlays the π* region. Both curves merge at higher
energies, reflecting the common spin and valence state (see
Figure 4. Ferrous cyt c: (a) Full 1s2p RIXS plane, (b) rising-edge
subtracted 1s2p RIXS plane, (c) comparison of the CEE cut through the
uncorrected plane (red line in (a)) with the CEE cut of Fe(II)TPP-
(ImH)2, (d) comparison of the CIE cut along the red line in (b) with the
CIE cut of Fe(II)TPP(ImH)2.
2
2
2
Figure S4 in the Supporting Information). A direct comparison
of the CIE cuts for the Fe(II)TPP(ImH)2 and ferrous cyt c, is
presented in (d). Two characteristic changes in spectral features
are visible: Fe(II) cyt c has lower intensity in the shoulder at
710 eV, but higher intensity in the L2 relative to the L3-edge.
Ferrous Multiplet Simulations. For quantitative analysis, the
experimental data have been simulated using charge-transfer
multiplet calculations within the VBCI framework. (see
Supporting Information) The best fit to Fe(II)TPP(ImH)2 is
shown in Figure 5: (a) K-edge and K-edge-like CEE cut, (b) the
1s2p RIXS plane, and (c) the L-edge and L-edge-like CIE cut.
The corresponding parameter set is given in Table S1. The
simulations reproduce the experimental results well for all three
experimental methods. In line with the experiment, the simulated
1s2p RIXS cut is broader than the L-edge XAS simulation. The
peak in the L-edge XAS simulation at 710 eV is reduced to a
shoulder in the RIXS cut, and the ratio of the L3/L2-peak maxima
is lower in the 1s2p RIXS cut. The projection of the VBCI ground
2
2
2
state gives 70.4% dx −y , 77.2% dz metal, and 8.1% dπ* back-
bonding character (Table 1).
D
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
into their individual pathways as shown in Figure 6(b−e). Since
intermediate states of different symmetries do not interfere, the
full RIXS plane is given by a summation of all four pathways,
which are compared to the L-edge XAS spectral splittings in
Figure 6(a).
The Eu final states contribute at energies above and below the
main peak with the excitation through the z2 (a1g) orbital being
lower in energy than through the x2 − y2 (b1g) orbital, reflecting
the ligand field energy splitting of these orbitals with the z2 being
lower in energy. The 1s2p RIXS final state splitting however is
significantly larger than the ground state energy splitting
(≈ 1.1 eV as compared to ≈0.4 eV,29 respectively), which results
from the strong 2p3d repulsive interaction. The Eu core-hole final
state has mainly px, py character (pz is mixed into the wave
function through 2p spin−orbit coupling), which have a lower
spatial overlap with the z2 than with the x2 − y2 orbital, leading
to the increased energy splitting. This is a characteristic of 1s2p
RIXS, which is inaccessible through L-edge XAS due to the
large differences in the electric dipole matrix elements for
absorption.48 In direct L-edge XAS, the transition matrix element
2
2
2
px,y → dx −y is four times stronger than px,y → dz , leading to an
intensity distribution mainly at higher energies (Figure 6(a)).
The other two final state symmetries are energetically located
between the Eu states. Both the B2u and A2u core-hole final states
(Figure 3(b)) have mainly pz character, which now has stronger
spatial overlap with the z2 orbital. Thus, the crystal field and 2p
core-hole repulsion influences on the energy oppose each other
with the core-hole being dominant, which inverts the excitation
energy order relative to the Eu states.
Figure 5. Fe(III)TPP(ImH)2 simulations: (a) comparison of the CEE
cut along the red line in (b) with the simulated K-edge, (b) 1s2p RIXS
plane simulation, (c) comparison of the CIE cut along the vertical red
line in (b) with the direct L-edge simulation.
For ferrous cyt c, simulations of the K-edge, 1s2p RIXS, and
CIE cut comparisons are shown in Figure 7. All relative changes
are reproduced, however they are less pronounced than in the
experiment. Both K-edge simulations have equal peak widths,
in line with the experimental result. The CIE cuts along the red
line in the 1s2p RIXS plane in Figure 7(c) show slightly higher
intensity in the shoulder at 710 eV for the model complex, while
the L2 intensity is higher for the ferrous cyt c simulation.
From the group theory considerations in Figure 3(b), B2u final
states are forbidden in direct L-edge XAS. In addition, final states
with Eu symmetry can be reached through the z2 and x2 − y2
intermediate states that are split in energy due to the tetragonal
geometry. Thus, the Eu final states are also split in energy by the
crystal field splitting of the dσ orbitals. To identify all
contributions to the 1s2p RIXS plane, we separate the excitations
Table 1. Unoccupied Metal d-Character Derived from DFT Calculations Using the BP86 Functional with 0, 10, and 20% Hartree−
a
Fock (HF) for the Full Molecule, 20% HF for the Truncated Model, and VBCI Simulations
2
2
2
dx −y [%]
dz [%]
π* [%]
d(π) [%]
model
protein
model
protein
model
protein
model
protein
ferrous:
full molecules:
BP86
71.3
75.0
77.3
71.6
74.8
77.3
71.8
75.8
78.5
70.4
73.6
76.5
9.3
7.2
5.2
9.2
7.0
4.8
BP86 + 10% HF
BP86 + 20% HF
truncated models:
BP86 + 20% HF
VBCI simulations:
77.7
70.4
77.6
73.4
78.5
77.2
77.9
70.4
5.2
8.1
5.2
7.1
ferric:
full molecules:
BP86
65.5
67.3
69.0
66.2
68.0
68.9
65.0
66.5
68.0
62.2
63.7
65.3
2.5
1.6
1.1
3.3
1.9
1.2
86.4
90.9
92.6
82.1
87.7
91.4
BP86 + 10% HF
BP86 + 20% HF
truncated models:
BP86 + 20% HF
VBCI simulations:
69.1
68.8
68.9
71.5
67.8
66.7
65.7
62.3
1.1
2.8
1.2
2.5
93.4
75.0
91.4
71.0
a
Note that the full molecule cyt c is optimized with the Fe−S(met) distance constrained at 2.29 (Fe(II)) and 2.33 Å (Fe(III)), while the truncated
2
model is fully optimized, and that the dz metal d-character includes the influence of both axial ligands.
E
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 6. Comparison of the final state L-edge XAS simulation (a), and the four 1s2p RIXS pathways carrying most intensity (b−e). Pathways through
the intermediate state with B2g and Eg are omitted for clarity.
(Table 1). Thus, in going from Fe(II)TPP(ImH)2 to Fe(II) cyt c,
2
2
2
the dz metal character decreases, while the dx −y increases and
the π back-bonding character slightly decreases due to the
normalized wave function.
Ferric Spectroscopy. Ferric TPP-bisIm. Figure 8 presents
the experimental results of Fe(III)TPP(ImH)2 plotted as for the
ferrous complexes: Figure 8(a) shows the full 1s2p RIXS plane,
while Figure 8(b) is the rising-edge subtracted plane. As a clear
difference relative to the ferrous results, a well separated feature is
visible at 7111.2 eV incident energy and 706 eV energy transfer
in the L3-edge-like region (7111.2/718.6 eV in the L2-edge-like
region). The high intensity feature around 7112.9/708.1 eV
(7112.9/720.6 eV) is similar to the low-spin ferrous case. Figure
8(c) gives a comparison of the direct K-edge XAS to the CEE cut
along the diagonal red line in Figure 8(a). In K-edge XAS, the low
energy peak at 7111.2 eV appears as a weak shoulder, while two
peaks are well resolved in the CEE cut. This is an intrinsic
advantage of the two-dimensional plane as given by 1s2p RIXS
and not due to experimental resolution, since the CIE integrated
RIXS plane (which represents the K-edge) overlays with the
direct K-edge spectrum (see Figure S5).
One of the main advantages of 1s2p RIXS is the ability to select
a specific incident energy (i.e., a specific K-pre-edge transition)
and record the corresponding L-edge-like spectrum associated
with this pre-edge peak. When exciting with 7111.2 eV incident
energy, the low lying dπ hole (in the ferric d5 configuration) is
probed, while exciting at an incident energy of 7112.7 probes
dσ character in the L-edge-like spectrum. Note that due to the
weak 1s3d interaction, little mixing occurs and pre-edge peakinter-
pretation in terms of specific d orbital excitations is reasonable,47
which is not the case for 2p L-edge final states.30
Figure 7. Ferrous cyt c simulations: (a) comparison of the CEE cut
along the red line in (b) with the CEE cut of the model complex
simulation, (b) 1s2p RIXS plane simulation, (c) comparison of the CIE
cut along the vertical red line in (b) with the CIE cut of the model
complex simulation.
In Figure 8(d), the direct L-edge XAS data are compared to
two CIE cuts taken through the two features in the rising-edge
subtracted RIXS plane (Figure 8(b)) along the green and red
lines at 7111.2 (π) and 7112.9 eV (σ character), respectively.
In order to reproduce these spectral differences, the covalent
mixing of the dz orbital must be increased in cyt c (see Table S1).
2
This leads to a projection of the VBCI ground state of 73.4%
2
2
2
d
x −y , 70.4% dz metal, and 7.1% dπ* back-bonding character
F
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 9. Ferric cyt c: (a) Full 1s2p RIXS plane, (b) rising-edge
subtracted 1s2p RIXS plane, (c) comparison of the CEE cut through the
uncorrected plane (red line in (a)) with the CEE cut of Fe(III)TPP-
(ImH)2, (d) comparison of the CIE cuts along the green and red lines in
(b) with the corresponding CIE cuts of the ferric model complex.
Figure 8. Fe(III)TPP(ImH)2: (a) Full 1s2p RIXS plane, (b) rising-edge
subtracted 1s2p RIXS plane, (c) comparison of the CEE cut through
the uncorrected plane (red line in (a)) with the direct K-edge, (d)
comparison of the CIE cuts along the green and red lines in (b) with the
direct L-edge experimental data.
through the ferric Fe(III)TPP(ImH)2 RIXS plane (dashed lines).
Comparing the two CIE spectra associated with π excitation
(cuts along the green lines in Figures 8 and 9(b)), three ob-
servations can be made: (1) The peak position of the low energy
feature is the same at 706 eV; (2) the protein peak is slightly
lower in intensity; and (3) the shoulder at 708.2 eV is higher in
intensity. CIE cuts associated with σ excitation (cuts along the
red lines in Figures 8 and 9(b))) reveal two distinct differences:
(1) The Fe(III) cyt c CIE spectrum is shifted toward lower
energies by 0.3 eV at both the L3- and L2-edge; (2) the L3/L2 peak
intensity ratio is lower for Fe(III) cyt c. Note that the lower L3/L2
peak intensity ratio was also found for the Fe(II) complexes,
while the shift to lower energy of the σ excitation decay spectrum
is only observed in the ferric complex comparison.
Ferric Multiplet Simulations. In parallel to the ferrous case,
charge-transfer multiplet calculations within the VBCI frame-
work were performed to simulate the above Fe(III) experimental
data. Final parameters are given in Table S1. The Slater−Condon
parameters were reduced to 70% of the Hartree−Fock derived
values in the initial and to 60% in the final states due to covalency
effects. The difference in ground and final states reduction
While the low energy cut aligns well with the lowest energy
L-edge XAS feature, the higher energy cut has its maximum at a
lower transfer energy than the main peak of the L-edge XAS
spectrum, similar to what was observed above for the ferrous
TPP(ImH)2 complex.
Ferric cyt c. Figure 9(a) and (b) presents the 1s2p RIXS
results for ferric cyt c, Figure 9(a) gives the full 1s2p RIXS plane,
while Figure 9(b) is the rising-edge subtracted plane. Three main
differences relative to the ferric TPP(ImH)2 model complex RIXS
data (Figure 8) are observed: (1) The π feature at 7111.1 eV
incident energy is weaker for the protein; (2) the strong σ feature
around 7112.6 eV is shifted down in incident energy by 0.3 eV
toward the π region; and (3) the L2-like σ region at 7112.6/720.3 eV
is higher in intensity relative to the L3-like σ region.
The first two points are also visible in Figure 9(c) where the
two CEE cuts of Fe(III) cyt c and TPP(ImH)2 are compared.
The low energy peak is slightly less intense and the main peak is
shifted to lower incident energy.
In Figure 9(d), the L-edge-like cuts through the cyt c RIXS
plane (solid lines) are compared to the corresponding cuts
G
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
originates from the effect of the mixing parameters on the
electron distribution in the VBCI model with and without a 2p
core-hole that are thus corrected using the Slater−Condon
parameters. Figure 10 gives the best fits to Fe(III)TPP(ImH)2
Figure 11. Ferric cyt c simulations: (a) Comparison of the CEE cut
along the red diagonal line in (b) with the CEE cut of the model complex
simulation, (b) 1s2p RIXS plane simulation, (c) comparison of the CIE
cuts along the green and red vertical lines in (b) with the CIE cuts of the
model complex simulation.
Figure 10. Fe(III)TPP(ImH)2 simulations: (a) Comparison of the CEE
cut along the red diagonal line in (b) with the simulated K-edge, (b)
1s2p RIXS plane simulation, (c) comparison of the CIE cuts along the
green and red vertical lines in (b) with the direct L-edge simulation.
plane, and (c) a comparison of the CIE cuts at lowest energy
through the dπ regions (along the green vertical lines in
Figures 11 and 10(b)), and the cuts at higher energies through
the dσ regions (along the red vertical lines in Figures 11 and 10(b)).
The CEE cuts of the simulated RIXS planes for both com-
plexes reproduce the experimental spectral changes well: the
intensity of the dπ peak at 7111.2 eV is decreased for cyt c, while
the energy difference between the dπand the dσ peak at ≈7113.0 eV
is 0.2 eV smaller in the ferric cyt c simulations, in line with the
experimental results in Figure 9.
In the L-edge-like CIE cuts through the simulated RIXS planes
of Fe(III)TPP(ImH)2 and ferric cyt c (Figure 11(c)), three main
changes in the spectral shape are observed: (1) A slight decrease
of the low energy Fe(III) cyt c peak intensity at 706 eV with an
increase in intensity in the shoulder at 708.2 eV in the CIE cut
after π excitation (green curves); (2) a shift by 0.3 eV toward
lower energy for the CIE cuts after σ excitation for ferric cyt c (red
curves); and (3) an increase of the L2-like σ intensity. All of these
characteristic changes are also observed in the CIE cuts through
the experimental RIXS planes (Figure 9(d)).
spectra: (a) The K-edge and K-edge-like CEE cut, (b) the 1s2p
RIXS plane, and (c) the L-edge and L-edge-like CIE cut. The
simulations reproduce the experimental results (see Figure 8) for
all three methods. The CEE cut along the diagonal red line in
Figure 10(b) (see Figure 10(a)) shows a higher intensity for the
dπ peak at 7111.2 eV compared to the direct K-edge simulation
due to the higher intrinsic resolution of the two-dimensional
RIXS plane. The dσ peak at ≈7113.2 eV appears at 0.3 eV higher
energy in the CEE cut than the direct K-edge simulation, in line
with the experimental result in Figure 9.
The direct L-edge simulation shown in Figure 10(c)
reproduces the experimental L-edge, but underestimates the
intensity of the lower energy peak at 706 eV. The dπ and dσ CIE
cuts (green and red vertical lines in Figure 10(b), respectively)
follow the experimental results, with the dπ intensity being
slightly overestimated, while the peak width, shape, and energy
splitting match the experimental results. The main CIE peak at
708.2 eV appears ∼1 eV lower in energy than the L3-edge peak,
similar to the behavior experimentally observed for Fe(II)TPP-
(ImH)2 above and in ferric complexes.38 A DOC quantitative
Similar to the ferrous case, these changes required an increase
in the covalent mixing of the z2 orbital for ferric cyt c relative to
Fe(III)TPP(ImH)2, together with a lower crystal field (10Dq)
and an increased covalent dπ mixing. This leads to a projection
2
2
2
2
2
2
analysis of the VBCI ground state reveals 69% dx −y , 67% dz , 75%
dπ metal character, and 3% dπ* back-bonding into the porphyrin.
In Figure 11, best fit simulations of ferric cyt c data are com-
pared to the above simulation results for Fe(III)TPP(ImH)2: (a)
A comparison of the CEE cut of ferric cyt c (along the diagonal
red line in Figure 11(b)) with the CEE cut of Fe(III)TPP(ImH)2
(along the diagonal red line in Figure 10(b)), (b) the 1s2p RIXS
analysis of the VBCI ground state DOC of 69% dx −y , 64% dz ,
73% dπ metal, and 3% dπ* back-bonding character in the
porphyrin. Thus, in going from ferric FeTPP(ImH)2 to cyt c, the
2
dz and dπ metal character decrease with the increase of the
2
2
covalent mixing. As a consequence, the dx −y character increases
and the π back-bonding character slightly decreases due to the
normalized wave function.
H
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Table 2. Collection of Metal and Ligand Characters from a Fragment Analysis for Three Different Functionals for the Full
a
Molecules as Well as for the Small Molecule Models
BP86
BP86 + 10% HF
BP86 + 20% HF
small molecule
model
protein
model
protein
model
protein
model
protein
ferrous:
dz2
71.8
9.7
16.7
−
70.4
6.5
75.8
8.4
14.3
−
73.6
6.0
78.5
7.3
11.6
−
76.5
4.1
78.5
7.4
11.8
−
77.9
5.0
por
imh2/met + imh
imh
20.6
10.8
9.8
18.3
9.6
13.8
7.7
16.3
8.4
met
2
−
−
8.7
−
6.1
−
7.9
ferric: dz
2
dz
65.0
10.3
19.7
−
62.2
8.9
66.5
10.4
19.2
−
63.7
8.7
68.0
9.9
18.4
−
65.3
8.1
67.8
11.1
18.7
−
65.7
7.9
por
imh2/met + imh
24.8
12.5
12.3
23.9
12.0
12.0
24.1
11.6
12.5
22.9
11.2
11.7
imh
met
−
−
−
−
ferric: dπ hole
dπ
86.4
13.3
2.9
−
82.1
18.7
3.1
90.9
9.1
2.7
−
87.7
12.8
1.8
92.6
6.7
2.2
−
91.4
10.7
1.4
93.4
6.9
2.4
−
91.4
9.8
0.0
0.0
0.0
por
imh2/met + imh
imh
met
3.1
1.8
1.4
−
0.0
−
0.0
−
0.0
−
a
Note that the full molecule cyt c is optimized with the Fe−S(Met) distance constrained, while the small molecule is fully optimized.
2
Density Functional Calculations. Correlating Electronic
A collection of the various ligand characters for dz binding as
derived from a fragment analysis of the DFT results is given
in Table 2 (top). All three functionals give the same behavior.
Structure Calculations to 1s2p RIXS Data. As described above,
VBCI multiplet calculations reproduce all observed spectro-
scopic changes in the 1s2p RIXS planes when going from both
the ferrous and ferric FeTPP(ImH)2 model complex to cyt c.
Both the ferrous and ferric complexes require an increase in the
2
The decrease in dz character is accompanied by a decrease in
porphyrin character and an increase in axial ligand character. This
is mainly due to the exchange of an imidazole by a thioether,
resulting in the ruffling of the porphyrin (the removing of the
cross-links do not significantly contribute to the ruffling50 and a
decrease in the Fe−N(His) bond length).
2
covalent a1g mixing for cyt c resulting in lower dz character (i.e.,
increased covalency), while for the dπ hole present in the ferric
case, the eg mixing increases for Fe(III) cyt c, leading to a decrease
in dπ character.
While VBCI simulations are able to reproduce the experi-
mental results due to the full multiplet treatment including the 1s
and 2p core-holes, they are unable to distinguish between the two
axial ligands and the specific contributions from the porphyrin, as
all contribute to the a1g (dz2) mixing.
DFT calculations were performed to obtain a more detailed
description of the ground state. The Fe−S(Met) distance in cyt c
was constrained to the crystal structure values of 2.29 Å (Fe(II)
and 2.33 Å (Fe(III)) since a fully optimized structure leads to an
overestimated Fe−S(Met) bond length by more than 0.1 Å.
However, also the fully optimized structure find a shorter
Fe(II)-S axial bond length, in-line with the experimental
multishell EXAFS finding16 and previous DFT results.49 The
resulting DFT derived charge densities (orbital differentiated
metal d-characters) are equivalent to the VBCI derived DOC
values and are used to couple the experimental results to the DFT
calculations.
The results of the DFT calculations for the ferric complexes
for all three functionals are given in Table 1 (bottom). While
2
2
both the dx −y and π* back-bonding covalencies do not change
between ferric cyt c and Fe(III)TPP(ImH)2, a decrease in dz2
metal character is found, similar to the ferrous case. The dπ hole
that is present in the ferric complexes also shows a decrease in
metal character for all functionals as found in the VBCI
simulations of the 1s2p RIXS data.
The fragment analysis (Table 2 (bottom)) qualitatively gives
the same ligand bonding behavior for the dz orbital in the ferric
as in the ferrous case: a decrease in dz and porphyrin ligand
character is compensated by an increase in axial ligand character
due to the exchange of an imidazole with the more covalent
thioether ligand. The dπ bonding is dominating between the dπ
hole and the filled porphyrin 3eg orbital, where the decrease in
metal dπ character reflects an increase in 3eg that is caused by the
ruffling of the porphyrin ring upon the imidazole/thioether
exchange. The imidazole bonding character in the dπ* hole is
small and unchanged between the Fe(III)TPP(ImH)2 and
Fe(III) cyt c and there is no Met contribution present in the π*
hole in ferric cyt c.
Truncated models with only the Fe-porphyrin ring and two
axial ligands (bis-His and His-Met) were geometry optimized
and compared to the results of the larger calculations. The last
column in Table 2 gives the results on the truncated models for
20% HF. Importantly, the Fe covalencies and fragment analyses
of the small molecule models quantitatively reproduce those
of the large molecule models that contain the propionates and
thioether linkages using the same functional and basis set. The
2
2
The results for the ferrous calculations are shown in Table 1
(top) for three functionals: BP86, BP86 + 10% Hartree−Fock
(HF), and BP86 + 20% HF (in the following labeled as
2
2
B(XXHF)P86). All three functionals find unchanged dx −y metal
d-character in ferrous cyt c compared to Fe(II)TPP(ImH)2 and a
2
small, reproducible decrease in dz character. The π* bonding
character does not change. These results parallel the VBCI
simulations of the data, where an increase in a1g mixing leads to a
2
2
2
decrease in dz character. The associated change in dx −y and π*
back-bonding in the VBCI simulations are due to the normalized
nature of the wave function.
I
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
a
Table 3. Thermodynamics of Ligand Loss from FeII/III
b
c
B(20HF)P86
FeII-S(Met)
FeIII-S(Met)
ΔΔ S(Met)
FeII-N(His)
FeIII-N(His)
ΔΔN(His)
FeII-H2O
FeIII-H2O
ΔΔH2O
ΔE
ΔH
ΔG
5.0
7.1
2.1
9.9
16.9
14.7
−0.4
4.9
7.0
7.5
8.8
0.2
−0.5
−2.0
−12.3
6.7
7.2
7.3
8.1
2.6
5.5
5.5
7.2
5.3
−13.9
3.6
−8.8
0.6
5.1
−9.1
4.7
−4.2
d
exp
−3.0
5.1
e
exp corr.
3.6
8.7
4.7
13.0
8.3
a
b
c
All energies are in kcal/mol. 6-311+G(d,p) basis set for all atoms and PCM corrections (ε = 4.0). ΔΔ is the difference in FeIII and FeII axial ligand
bond strengths. ΔG0 values obtained from dissociation constants for AcMet and imidazole.10 Experimental values corrected for the calculated
ΔΔG of H2O dissociation from Fe(III) (i.e., 8.1 kcal/mol)
d
e
small molecule models are thus used below to quantitatively
evaluate the thermodynamics of axial ligand metal bonding.
Axial Ligand Bond Strengths. In this section, axial ligand
bond strengths are calculated for both N(His) and S(Met) axial
ligands in both Fe(II) and Fe(III) oxidation states. The bond
strength is taken as the difference in energy between the ligand-
on and ligand-off forms. Calculations do not include dispersion
corrections; however, this should mainly affect the absolute
binding energies and not the relative bond strengths.
A range of functionals and hybrids was evaluated to best
reproduce the proper spin ground state for ligand-on and ligand-
off forms for both redox states. This is described in the
Supporting Information. In the presentation below, we use the
B(20HF)P86 hybrid functional. However, all functionals lead to
equivalent ligand binding descriptions.
The calculated axial ligand bond strengths for S(Met) and
N(His) in both Fe(II) and Fe(III) oxidation states are given in
Table 3. Note that values are reported for ligand loss. The axial
ligand Fe(II)-S(Met) bond strength is calculated to be 5.0 kcal/mol
(ΔE), while the Fe(II)-N(His) bond strength is 9.9 kcal/mol; the
axial ligand Fe(III)-S(Met) bond strength is 7.1 kcal/mol, while
the Fe(III)-N(His) bond strength is 16.9 kcal/mol. Thus, the
Fe−S(Met) bond is weaker than the Fe−N(His) in both Fe
oxidation states. Interestingly, from the above 1s2p RIXS data
and the differential orbital covalency, the Fe−S(Met) bond is
more covalent than the Fe−N(His) bond. This apparent discrep-
ancy between covalencies and bond strengths is addressed in the
Discussion. Furthermore, the Fe(III)-S(Met) bond is stronger
than the Fe(II)-S(Met) bond (ΔE = 7.1 vs 5.0 kcal/mol,
respectively). This is consistent with the higher S(Met) character
for Fe(III) cyt c from the RIXS data/simulations and DFT
calculations presented above. However, this observation is not
consistent with the literature considerations summarized in the
Introduction. Lastly, the difference in Fe(III) and Fe(II) bond
strengths is larger for a N(His) axial ligand than for S(Met)
(Table 3, ΔΔE = 7.0 vs 2.1 kcal/mol, respectively; note the ΔΔG
for H2O is also given in Table 3 and is referred to in the
Discussion). The stronger bonds to the Fe(III) states are
generally consistent with this higher degree of covalency than in
the corresponding Fe(II) states.
the surface of the sample. The sampling depth can be enlarged to
a few micrometers by detecting the fluorescence decay; however,
this can lead to significant self-absorption effects. Hard X-rays do
not have these restrictions, but give lower resolution spectra due
to the short lifetime of the 1s core-hole as observed in Fe
K-edges.47
1s2p RIXS combines the advantages of both methods, leading
to two-dimensional high resolution spectra for samples in a
variety of environments. The same information regarding the
DOC can be extracted from 1s2p RIXS with the advantage of
accessing L-edge XAS forbidden final states.38 In complexes with
tetragonal symmetry like FeTPP(ImH)2, where both L-edge
XAS and 1s2p RIXS have been obtained, the broadening of the
L-edge-like (CIE) RIXS cut associated with σ excitation is further
2
2
2
enlarged due to the contribution from the dx −y −dz orbital
energy splitting, while this equatorial vs axial splitting is not
accessible in direct L-edge XAS due to the dipole selection rule
and strong multiplet effects.
To access the electronic structure of the heme center in cyt c in
its ferrous and ferric form, only 1s2p RIXS experiments could
be performed; these were compared to the results from cor-
responding ferrous and ferric FeTPP(ImH)2 model complexes.
In the ferrous case, two characteristic spectral changes were
observed for ferrous cyt c relative to the ferrous bis-imidazole
model: A decrease in intensity of the high energy shoulder of
the L3-like peak and an increase of relative intensity of the L2-like
peak in the CIE cut associated with σ excitation (Figure 4). These
changes require an increase of the z2 covalency. Note that
these spectral changes can only be quantified by including the
whole L-edge-like energy range; often, only the L3-like energy
region is taken into account, which would not have been
sufficient for a unique identification of this increase in z2
covalency. In the ferric complexes, two CIE cuts are required
and their comparison between the ferric bis-imidazole model and
ferric cyt c defines four characteristic spectral variations: (i) The
L2-like relative intensity increases for ferric cyt c in the CIE RIXS
cut associated with σ excitation, similar to the ferrous case; (ii)
the energy positions of the L3- and L2-like peaks in the σ cut are
shifted toward lower energies, which is different from the ferrous
result; (iii) the main peaks in the CIE cut associated with π
excitation appear at the same energy, but with weaker intensity
for ferric cyt c; and (iv) the intensity of the shoulder of the main
peak in the π cut increases (Figure 8). An increase of z2 covalency
reproduces the increase in L2-like intensity, parallel to the ferrous
case. The shift toward lower energy of the σ CIE cut is due to a
decrease of the crystal field strength, 10Dq, together with an
increase of π covalency, reproducing the other three observa-
tions. From DFT calculations correlated to these data, the
increase of z2 covalency is due to the axial ligand change from
imidazole to methionine, as methionine has a stronger σ dona-
tion than imidazole. The increase of π covalency is due to the
DISCUSSION
■
The highly covalent environment in Fe porphyrin complexes do
not allow for most methods to quantitatively determine the local
electronic structure. Fe L-edge XAS using soft X-rays is a
powerful tool to determine the metal d-characters for all orbitals
(i.e., the DOC)27 including the ability to separately determine
the σ and π donation and back-bonding.28,29,51 The soft X-ray
nature of Fe L-edge XAS leads to a low lifetime broadening and a
feature rich spectrum, but requires ultrahigh vacuum conditions
and generally involves the detection of electrons within a few Å of
J
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
porphyrin ring, which is ruffled due to this ligand exchange,
rather than a π donation interaction with the methionine ligand.
Finally, from the quantitative DOC’s obtained from the VBCI fit
to the 1s2p RIXS data, the axial methionine covalent interaction
is found to be larger in ferric than in ferrous cyt c.
Fe(II)-S(Met) bond strength). The energy of ligand loss on the
S = 0 surface is calculated to be 12.3 kcal/mol (Figure 12, right).
The difference in energy for these two processes of ligand loss is
the exchange stabilization energy corresponding to a change in
Fe(II) spin-state (i.e., S = 0 → S = 2) and is calculated to be
6.2 kcal/mol (Figure 12). (Note that the exchange stabilization
energy is taken to be the difference in energy between the S = 0
and S = 2 spin states without an additional axial ligand in order to
directly compare to Fe(III) case below.) Thus, this exchange
stabilization lowers the calculated value for the axial ligand bond
strength. Analogous calculations for axial ligand loss from the
Fe(III)-porphyrin site give gas-phase axial ligand bond strengths
of 8.1 kcal/mol (for the S = 1/2 → S = 3/2 conversion) and
17.8 kcal/mol (for ligand loss along the S = 1/2 surface) and an
exchange stabilization energy of 9.7 kcal/mol. Thus, even after
correcting the Fe(III) and Fe(II) axial ligand bond strengths for
the different exchange stabilization energies (9.7 vs 6.2 kcal/mol,
respectively), the axial ligand bond in the Fe(III) site is still
stronger than that for the Fe(II) (8.1 vs 5.6 kcal/mol, respec-
tively). Without exchange stabilization, the difference between
the Fe(III) and Fe(II) bond strengths is even larger (17.8 vs
12.3 kcal/mol). The fact that the Fe(III)-S(Met) bond is
stronger than Fe(II)-S(Met) is consistent with the higher degree
of covalency, as determined above via the 1s2p RIXS data and
simulations.
As outlined in the Introduction, past considerations have
concluded that, in cyt c, the Fe(II)-S(Met) bond is stronger
than the Fe(III)-S(Met) bond. One argument for this is the
shorter Fe(II)-S(Met) bond (2.29 Å) relative to Fe(III)-S(Met)
(2.33 Å) bond.16 However, the bond lengths are not necessarily
related to bond strengths, as these are affected differently by the
ionic and covalent contributions to bonding.30,53 Furthermore,
for considerations related to the decrease in E0 upon Fe−S(Met)
bond loss,6,7 ligand competition8,9 and binding,10−12 and protein
folding studies,10,13−15 it is essential to define and refer to the
reference state (i.e., S(Met) bound vs either N(His), N(Lys), or
H2O axial ligands). Importantly, from Table 3, the ΔΔG for axial
ligand binding between Fe(III) and Fe(II) states for S(Met),
N(His), and H2O axial ligands is 5.1, 8.8, and 8.1 kcal/mol,
respectively, and is thus larger for the N(His) and H2O axial
ligands. This difference results in a decrease in E0 upon loss of the
S(Met) ligand and replacement with either a nitrogen or oxygen
based ligand (estimated to be 160 mV from the ΔΔG for S(Met)
vs N(His), Table 3). This difference in ΔΔG also leads to an
apparent increase in S(Met) binding affinity to Fe(II) relative to
Fe(III),10−12 as binding to the Fe(III) (but not Fe(II)) state also
involves displacement of an axial H2O ligand. The calculated
thermodynamics of ligand loss can be compared to experimental
data obtained from the dissociation constants of AcMet and
imidazole measured for N-acetylmicroperoxidase-8 (AcMP8, see
Table 3).10 The ΔG0s of AcMet ligand loss from Fe(II) and
Fe(III) AcMP8 are 3.6 and 0.6 kcal/mol, respectively. This
would give an apparent ΔΔG of −3.0 kcal/mol, with AcMet
binding stronger to Fe(II). However, correcting for water
binding/displacement from Fe(III) and not Fe(II) using the
calculated ΔΔG of 8.1 kcal/mol (Table 3) gives an adjusted
ΔΔG of 5.1 kcal/mol. This compares well to the calculated value
of 5.1 kcal/mol as given in Table 3. The same analysis can be
applied to an imidazole axial ligand; this results in an adjusted
ΔΔG of 8.3 kcal/mol, which compares well to the calculated
value of 8.8 kcal/mol. The difference in the ΔΔG for S(Met) vs
N(His) or H2O ligands can be understood through hard/soft
acid/base concepts, where the harder nitrogen- and oxygen-based
The experimental RIXS data and simulations have indicated
that the Fe−S(Met) bond is more covalent than the Fe−N(His)
bond. However, the calculated axial ligand bond strengths in
Table 3 indicated that the Fe−S(Met) bond is weaker than the
Fe−N(His) bond. This difference can be understood in terms of
basic bonding concepts. Within the framework of perturbation
theory, the bond energy (BE) is proportional to (HM−L)2/Δ,
where HM−L is the resonance integral between the metal and the
ligand orbitals and Δ is the difference in energy between these
interacting orbitals before bonding. The covalency (i.e., the
coefficient squared of ligand character in the metal d-orbital, α2)
is proportional to ((HM−L)/Δ)2. Thus, the BE can be estimated
from the covalency scaled by Δ (i.e., BE = α2Δ). Thus, the
stronger Fe−N(His) bonds relative to Fe−S(Met), despite their
lower covalencies, can be understood in terms of the difference in
their valence donor orbital energies, where the ImH σ donor
bond is 1.1 eV lower in energy than the Met b1 valence donor
orbital. This results in a significantly larger Δ for N(His) and thus
a stronger M−L bond.52
It was found above that the Fe(III)-S(Met) bond is stronger
than the Fe(II)-S(Met) bond (7.1 vs 5.0 kcal/mol, respectively,
Table 3). The energy of the axial ligand bond was calculated as
the difference in energy between the ligand-on (S = 1/2 and S = 0
for Fe(III) and Fe(II), respectively) and ligand-off ground states
(S = 3/2 quantum mixed and S = 2 for Fe(III) and Fe(II),
respectively). Thus, there can be additional contributions to the
calculated bond strengths due to the different spin states involved
in ligand loss. These can be taken into account using the potential
energy surfaces (PESs) of ligand binding to the Fe(III) and
Fe(II) states. Figure 12 gives the PESs for Met ligand loss for the
Figure 12. DFT calculated gas-phase potential energy surfaces of
S(Met) binding to both Fe(II) S = 0 (black line and crosses) and S = 2
(red dashed line and crosses) spin-states. Values for the energetics are
given in kcal/mol. Energetics are calculated from the ligand-off form
either with the porphyrin and methionine in the same calculation or
porphyrin and methionine in separate calculations (in parentheses).
Values for Fe(III) S = 1/2 and S = 3/2 are given in the text.
ferrous state in the gas phase leading to an S = 0 to S = 2 surface
crossing. The gas-phase axial S(Met) ligand bond strength is
5.6 kcal/mol. Note that energies for the corresponding two-box
calculations (i.e., the prophyrin and axial ligand are calculated
separately) are given in parentheses (e.g., 5.8 kcal/mol for the
K
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
ligands bind more strongly to the Fe(III) than Fe(II) oxidation
state and while the softer sulfur-based ligand still binds more
strongly to Fe(III), it has a higher relative (to N and O based
ligands) affinity for Fe(II).
information on spectral comparisons for the ferric compounds.
Further details on the ligand bond strength calculations are given
together with detailed tables. This material is available free of
charge via the Internet at http://pubs.acs.org.
Previous studies have defined the function of the axial S(Met)
ligand in T1 Cu and CuA.2,4,5 The presence of the axial S(Met)
bond lowers E0 due to the stronger Cu(II)-S(Met) bond relative
to Cu(I). This difference in E0 is relative to the absence of axial
ligand binding upon S(Met) ligand loss. A similar function is
found here for cyt c, as the Fe(III)-S(Met) bond is indeed
stronger than Fe(II); however, decreases in E0 are observed upon
ligand loss in cyt c due to its replacement by an endogenous
(N(His) or N(Lys)) or exogenous ligand (H2O). These ligands
have larger ΔGs for Fe(III) vs Fe(II) binding and thus effec-
tively lower E0 relative to the S(Met). Furthermore, the T1
Cu−S(Met) bond has been measured experimentally to be
weak.5 Thus, the S(Met) bond would be subject to ligand loss
due to entropic contributions at physiological temperature.4,5
This has provided insight into the “entatic/rack”54−57 nature of
T1 Cu and CuA ET active sites. Similar behavior is found here for
cyt c, as the axial ligand bond strengths for both Fe(II) and
Fe(III) are weak (i.e., ∼5 kcal/mol) and on the order of the TΔS
free energy of bond loss. Thus, analogous to T1 Cu, the protein
matrix of cyt c provides the opposing free energy necessary to
keep the axial S(Met) bond intact. The axial Fe−S(Met) bond in
cyt c can therefore be considered under entatic control. The
weaker Fe(II)-S(Met) bond as compared to the Fe(III)-S(Met)
bond is also consistent with the observed axial ligand photolysis
for Fe(II) but not Fe(III) cyt c. Differentiating bond strength
contributions to the photochemistry and photophysics in cyt c is
an aim of current studies.
AUTHOR INFORMATION
■
Corresponding Authors
hedman@slac.stanford.edu
hodgson@slac.stanford.edu
edward.solomon@stanford.edu
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
T.K. acknowledges financial support by the German Research
Foundation (DFG) grant KR3611/2-1. R.G.H. acknowledges a
Gerhard Casper Stanford Graduate Fellowship and the Achieve-
ment Rewards for College Scientists Foundation (ARCS). M.L.
acknowledges the Marcus and Amalia Wallenberg foundation and
the Swedish Research Council. This research was supported by the
National Institute of General Medical Sciences of the National
Institutes of Health under award number R01GM040392 (E.I.S.),
NSF grant CHE-1360046 (E.I.S.) and P41GM103393 (K.O.H).
Use of the Stanford Synchrotron Radiation Lightsource, SLAC
National Accelerator Laboratory, is supported by the U.S.
Department of Energy, Office of Science, Office of Basic Energy
Sciences under Contract No. DE-AC02-76SF00515. The SSRL
Structural Molecular Biology Program is supported by the DOE
Office of Biological and Environmental Research, and by the
National Institutes of Health, National Institute of General
Medical Sciences (including P41GM103393). Parts of this work
was performed at the Linac Coherent Light Source (LCLS), SLAC
National Accelerator Laboratory, which is supported by the U.S.
Department of Energy, Office of Science, Office of Basic Energy
Sciences under Contract No. DE-AC02-76SF00515. This research
used resources of the Advanced Photon Source, a U.S.
Department of Energy (DOE) Office of Science User Facility
operated for the DOE Office of Science by Argonne National
Laboratory under Contract No. DE-AC02-06CH11357. We
acknowledge Ryan Wilson for the kind provision of computer
power, and Matthew Kieber-Emmons for help in the purification
and preparation of cyt c samples.
CONCLUSION
■
This study has employed Fe 1s2p RIXS to investigate the
electronic structure of ferrous and ferric cyt c. New insights were
obtained by correlating cyt c to a bis-imidazole porphyrin model
complex to experimentally define the differential orbital
covalency of the iron site in the highly covalent porphyrin environ-
ment. They indicate an increased covalency for the Fe−S(Met)
axial bond relative to Fe−N(His) as well as a higher degree of
covalency for the ferric relative to the ferrous state. DFT
calculations correlated to data further allowed for the evaluation
of the relative axial ligand bond strengths for S(Met) and N(His)
in both Fe(II) and Fe(III) redox states. Importantly, for cyt c,
these calculations indicate that the Fe−N(His) bonds are
stronger than the Fe−S(Met) bonds despite the latter being
more covalent. Furthermore, the Fe(III)-S(Met) bond is
stronger than the Fe(II)-S(Met) bond, which provides new
insight into the redox and photophyscial properties of this
protein. Finally, both (Met)S−Fe bonds are weak and an
important role of the protein is in maintaining these bonds under
physiological conditions.
REFERENCES
■
(1) Liu, J.; Chakraborty, S.; Hosseinzadeh, P.; Yu, Y.; Tian, S.; Petrik, I.;
Bhagi, A.; Lu, Y. Chem. Rev. 2014, 114, 4366.
(2) Solomon, E. I.; Szilagyi, R. K.; DeBeer, G. S.; Basumallick, L. Chem.
Rev. 2004, 104, 419.
(3) Solomon, E. I.; Hadt, R. G. Coord. Chem. Rev. 2011, 255, 774.
(4) Tsai, M.-L.; Hadt, R. G.; Marshall, N. M.; Wilson, T. D.; Lu, Y.;
Solomon, E. I. Proc. Natl. Acad. Sci. U. S. A. 2013, 110, 14658.
(5) Ghosh, S.; Xie, X.; Dey, A.; Sun, Y.; Scholes, C. P.; I, S. E. Proc. Natl.
Acad. Sci. U. S. A. 2009, 106, 4969.
ASSOCIATED CONTENT
* Supporting Information
■
(6) Raphael, A. L.; Gray, H. B. Proteins: Struct., Funct., Bioinf. 1989, 6,
S
338.
Additional information to the molecular structures are given and
detailed information on sample preparation for both FeTPP-
(ImH)2 and cyt c are presented. The experimental methods
L-edge XAS, K-edge XAS and 1s2p RIXS are described in detail,
as well as the theoretical methodology used regarding the charge
transfer multiplet calculations, plus a detailed description of the
calculations using DFT. The methodology part of the Results
and Analysis section is presented here together with additional
(7) Ye, T.; Kaur, R.; Senguen, F. T.; Michel, L. V.; Bren, K. L.; Elliott, S.
J. J. Am. Chem. Soc. 2009, 130, 6682.
(8) Hantgan, R. R.; Taniuchi, H. J. Biol. Chem. 1978, 253, 5373.
(9) Schejter, A.; Plotkin, B. J. Biol. Chem. 1988, 255, 353.
(10) Tezcan, F. A.; Winkler, J. R.; Gray, H. B. J. Am. Chem. Soc. 1998,
120, 13383−13388.
(11) Vashi, P. R.; Marques, H. M. J. Inorg. Biochem. 2004, 98, 1471−
1482.
L
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(12) Lushington, G. H.; Cowley, A. B.; Silchenko, S.; Lukat-Rodgers,
G. S.; Rodgers, K. R.; Benson, D. R. Inorg. Chem. 2003, 42, 7550−7559.
(13) Pascher, T.; Chesick, J. P.; Winkler, J. R.; Gray, H. B. Science 1996,
271, 1558−60.
(41) Gog, T.; Casa, D. M.; Said, A. H.; Upton, M. H.; Kim, J.;
Kuzmenko, I.; Huang, X.; Khachatryan, R. J. Synchrotron Radiat. 2013,
20, 74−79.
(42) Cowan, R. D. The Theory of Atomic Structure and Spectra;
University of California Press: Berkeley, 1981.
(14) Mines, G. A.; Pascher, T.; Lee, S. C.; Winkler, J. R.; Gray, H. B.
Chem. Biol. 1996, 3, 491−497.
(43) Butler, P. H. Point Group Symmetry: Applications, Methods and
Tables; Plenum Press: New York, 1981.
(15) Cohen, D. S.; Pielak, G. J. J. Am. Chem. Soc. 1995, 117, 1675−7.
(16) Cheng, M.-C.; Rich, A. M.; Armstrong, R. S.; Ellis, P. J.; Lay, P. A.
Inorg. Chem. 1999, 38, 5703.
(44) Thole, B. T.; van der Laan, G.; C.Fuggle, J.; Sawatzky, G. A.;
Karnatak, R. C.; Esteva, J.-M. Phys. Rev. B: Condens. Matter Mater. Chem.
1985, 32, 5107−5118.
(45) de Groot, F. M. F.; Kotani, A. Core Level Spectroscopy of Solids;
CRC Press: Boca Raton, FL, 2008.
(17) In the literature, there is a range of Fe-thioether bond length of
multiple cyt c systems given. The resolution of the diffraction data is not
sufficient to distinguish between Fe(II)- and Fe(III)-thioether bond
lengths. EXAFS data have the precision, but the result depends on the
number of shells that are taken into account. Single-shell EXAFS analysis
results in more similar bond lengths, while multishell analysis finds a
clear separation between them with Fe(II)-S being shorter.
(18) Liu, X.; Kim, C. N.; Yang, J.; Jemmerson, R.; Wang, X. Cell 1996,
86, 147−157.
(19) Kagan, V. E.; et al. Nat. Chem. Biol. 2005, 1, 223−232.
(20) Belikova, N. A.; Jiang, J.; Tyurina, Y. Y.; Zhao, Q.; Epperly, M. W.;
Greenberger, J.; Kagan, V. E. Int. J. Radiat. Oncol. Biol. Phys. 2007, 69,
176−186.
(21) Kagan, V. E.; Bayir, H. A.; Belikova, N. A.; Kapralov, O.; Tyurina,
Y. Y.; Tyurin, V. A.; Jiang, J.; Stoyanovsky, D. A.; Wipf, P.; Kochanek, P.
M.; Greenberger, J. S.; Pitt, B.; Shvedova, A. A.; Borisenko, G. Free
Radical Biol. Med. 2009, 46, 1439−1453.
(22) Wang, W.; Ye, X.; Demidov, A. A.; Rosca, F.; Sjodin, T.; Cao, W.;
Sheeran, M.; Champion, P. M. J. Phys. Chem. B 2000, 104, 10789.
(23) Negrerie, M.; Cianetti, S.; Vos, M. H.; Martin, J.-L.; Kruglik, S. G.
J. Phys. Chem. B 2006, 110, 12766.
(46) Frisch, M. J. et al. Gaussian 09, Revision D.01; Gaussian, Inc.:
Wallingford, CT, 2009.
(47) Westre, T. E.; Kennepohl, P.; DeWitt, J. G.; Hedman, B.;
Hodgson, K. O.; Solomon, E. I. J. Am. Chem. Soc. 1997, 119, 6297−6314.
(48) Kroll, T.; Kraus, R.; Schonfelder, R.; Aristov, V. Y.; Molodtsova,
̈
O. V.; Hoffmann, P.; Knupfer, M. J. Chem. Phys. 2012, 137, 054306.
(49) Rovira, C.; Carloni, P.; Parrinello, M. J. Phys. Chem. B 1999, 103,
7031−7035.
(50) Faraone-Mennella, J.; Tezcan, F. A.; Gray, H. B.; Winkler, J. R.
Biochemistry 2006, 45, 10504−10511.
(51) Wilson, S. A.; Kroll, T.; Decreau, R. A.; Hocking, R. K.; Lundberg,
M.; Hedman, B.; Hodgson, K. O.; Solomon, E. I. J. Am. Chem. Soc. 2013,
135, 1124−1136.
(52) An equivalent explanation can be given in terms of the difference
in electronegativity of the donor atoms in the axial residues.
(53) Gorelsky, S. I.; Basumallick, L.; Vura-Weis, J.; Sarangi, R.;
Hodgson, K. O.; Hedman, B.; Fujisawa, K.; Solomon, E. I. Inorg. Chem.
2005, 44, 4947.
(54) Gray, H. B.; Malmstrom, B. G.; Williams, R. J. P. J. Biol. Inorg.
̈
(24) Zang, C.; Stevens, J. A.; Link, J. J.; Guo, L.; Wang, L.; Zhong, D. J.
Am. Chem. Soc. 2009, 131, 2846.
(25) Braem, O.; Consani, C.; Cannizzo, A.; Chergui, M. J. Phys. Chem.
B 2011, 115, 13723.
(26) Solomon, E. I.; Hedman, B.; Hodgson, K. O.; Dey, A.; Szilagyi, R.
Chem. 2000, 5, 551.
(55) Williams, R. J. P. Eur. J. Biochem. 1995, 234, 363.
(56) Malmstrom, B. G. Eur. J. Biochem. 1994, 223, 711.
̈
(57) Winkler, J. R.; Wittung-Stafshede, P.; Leckner, B. G. J.;
Malmstrom; Gray, H. B. Proc. Natl. Acad. Sci. U. S. A. 1997, 94,
̈
K. Coord. Chem. Rev. 2005, 249, 97.
4246−4249.
(27) Wasinger, E. C.; de Groot, F. M. F.; Hedman, B.; Hodgson, K. O.;
Solomon, E. I. J. Am. Chem. Soc. 2003, 125, 12894−12906.
(28) Hocking, R. K.; Wasinger, E. C.; de Groot, F. M. F.; Hodgson, K.
O.; Hedman, B.; Solomon, E. I. J. Am. Chem. Soc. 2006, 128, 10442−
10451.
(29) Hocking, R. K.; Wasinger, E. C.; Yan, Y.-L.; deGroot, F. M. F.;
Walker, F. A.; Hodgson, K. O.; Hedman, B.; Solomon, E. I. J. Am. Chem.
Soc. 2007, 129, 113−125.
(30) Hocking, R. K.; DeBeer George, S.; Raymond, K. N.; Hodgson, K.
O.; Hedman, B.; Solomon, E. I. J. Am. Chem. Soc. 2010, 132, 4006−4015.
̈
(31) Gel’mukhanov, F.; Ogreen, H. Phys. Rep. 1999, 312, 87−330.
(32) Kotani, A.; Shin, S. Rev. Mod. Phys. 2001, 73, 203−246.
(33) Glatzel, P.; Bergmann, U.; Yano, J.; Visser, H.; Robblee, J. H.; Gu,
W.; de Groot, F. M. F.; Christou, G.; Pecoraro, V. L.; Cramer, S. P.;
Yachandra, V. K. J. Am. Chem. Soc. 2004, 126, 9946−9959.
(34) de Groot, F. M. F.; Glatzel, P.; Bergmann, U.; van Aken, P. A.;
Barrea, R. A.; Klemme, S.; Havecker, M.; Knop-Gericke, A.; Heijboer,
̈
W. M.; Weckhuysen, B. M. J. Phys. Chem. B 2005, 109, 20751−20762.
(35) Glatzel, P.; Bergmann, U. Coord. Chem. Rev. 2005, 249, 65−95
Synchrotron Radiation in Inorganic and Bioinorganic Chemistry.
(36) Leidel, N.; Chernev, P.; Havelius, K. G. V.; Schwartz, L.; Ott, S.;
Haumann, M. J. Am. Chem. Soc. 2012, 134, 14142−14157.
(37) Glatzel, P.; Bergmann, U.; Yano, J.; Visser, H.; Robblee, J. H.; Gu,
W.; de Groot, F. M. F.; Christou, G.; Pecoraro, V. L.; Cramer, S. P.;
Yachandra, V. K. J. Am. Chem. Soc. 2004, 126, 9946−9959.
(38) Lundberg, M.; Kroll, T.; DeBeer, S.; Bergmann, U.; Wilson, S. A.;
Glatzel, P.; Nordlund, D.; Hedman, B.; Hodgson, K. O.; Solomon, E. I. J.
Am. Chem. Soc. 2013, 135, 17121−17134.
(39) Sokaras, D.; Weng, T.-C.; Nordlund, D.; Alonso-Mori, R.;
Velikov, P.; Wenger, D.; Garachtchenko, A.; George, M.; Borzenets, V.;
Johnson, B.; Rabedeau, T.; Bergmann, U. Rev. Sci. Instrum. 2013, 84,
053102.
(40) Gog, T.; et al. Synchrotron Radiat. News 2009, 22, 12−21.
M
dx.doi.org/10.1021/ja5100367 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX