Direct observation of oxygen rebound with an iron-hydroxide complex

Article abstract of DOI:10.1021/jacs.7b07979

The rebound mechanism for alkane hydroxylation was invoked over 40 years ago to help explain reactivity patterns in cytochrome P450, and subsequently has been used to provide insight into a range of biological and synthetic systems. Efforts to model the rebound reaction in a synthetic system have been unsuccessful, in part because of the challenge in preparing a suitable metalhydroxide complex at the correct oxidation level. Herein we report the synthesis of such a complex. The reaction of this species with a series of substituted radicals allows for the direct interrogation of the rebound process, providing insight into this uniformly invoked, but previously unobserved process.

Full text of DOI:10.1021/jacs.7b07979

Communication
pubs.acs.org/JACS
Direct Observation of Oxygen Rebound with an Iron-Hydroxide
Complex
Jan Paulo T. Zaragoza,† Timothy H. Yosca,‡ Maxime A. Siegler,† Pierre Moenne-Loccoz,§
̈
Michael T. Green,‡ and David P. Goldberg*,†
†Department of Chemistry, The Johns Hopkins University, 3400 N Charles Street, Baltimore, Maryland 21218, United States
‡Department of Chemistry and Department of Molecular Biology and Biochemistry, University of California, Irvine, California 92697,
United States
§Division of Environmental and Biomolecular Systems, Oregon Health & Science University, Portland, Oregon 97239, United States
S
* Supporting Information
species, Compound-I (Cpd-I, FeIV(O)(porph+•)), leading to
protonated Compound-II (Cpd-II, FeIV(OH)(porph)), fol-
ABSTRACT: The rebound mechanism for alkane hydrox-
ylation was invoked over 40 years ago to help explain
reactivity patterns in cytochrome P450, and subsequently
has been used to provide insight into a range of biological
and synthetic systems. Efforts to model the rebound
reaction in a synthetic system have been unsuccessful, in
part because of the challenge in preparing a suitable metal-
hydroxide complex at the correct oxidation level. Herein
we report the synthesis of such a complex. The reaction of
this species with a series of substituted radicals allows for
the direct interrogation of the rebound process, providing
insight into this uniformly invoked, but previously
unobserved process.
lowed by “rebound” of the newly formed metal-bound
hydroxide to the carbon radical to give the hydroxylated
product and the one-electron reduced FeIII(porph).1e,f
Efforts over the years have led to evidence that supports the
rebound mechanism, including studies with radical clock
probes,1a as well as myriad computational analyses.1g However,
direct observation of the radical rebound step has been
impossible because the initial C−H cleavage step is invariably
rate-determining, making rebound invisible to spectroscopic
interrogation. It was not until recently that one of us showed
Cpd-I in CYP could be characterized spectroscopically,1e which
did allow for direct study of the C−H cleavage step, but left the
radical rebound step unobserved.
One way to overcome the problems associated with
observing the rebound process is to independently prepare
the rebound intermediate and examine its reactivity with well-
defined organic radicals. However, this strategy has been
unsuccessful to date because of the inherent difficulties in
generating discrete, well-characterized Fe(OH) species at the
correct redox level.2 The putative rebound intermediate, i.e.,
protonated Cpd-II ((FeIVOH)(porph)), has predominantly
been trapped and spectroscopically observed in heme proteins
containing anionic axial ligands,3 leading to a trianionic
environment at the metal. To our knowledge, synthetic analogs
of protonated Cpd-II remain unknown. We hypothesized that
the corrole ligand, which presents a trianionic donor set
without the need for an axial donor, could act to stabilize the
key [Fe(OH)]3+ state. Herein we report the development of an
iron-hydroxide Cpd-II model complex, and demonstrate that it
can be used for the direct interrogation of radical rebound
reactions.
he selective hydroxylation of C−H bonds is a critical
Ttransformation for a wide range of biological processes,
including oxidative metabolism, the detoxification of xeno-
biotics, and the biosynthesis of natural products. The same
transformation is of widespread importance in organic
synthesis, from the preparation of pharmaceuticals, to the
synthesis of numerous specialty and commodity chemicals.
Both heme and nonheme metalloenzymes carry out C−H
hydroxylations with high efficiency and selectivity.1a,b In
comparison, there are relatively few synthetic complexes,
especially with biologically compatible metal ions, that can
mediate the same C−H hydroxylation chemistry.1c,d The
canonical mechanism invoked for C−H hydroxylation is the
so-called “rebound” mechanism, exemplified by cytochrome
P450 (CYP) (Scheme 1). The initial step involves H atom
abstraction from a C−H bond by a high-valent metal-oxo
Part of the synthetic challenge to prepare an appropriate
mononuclear Fe(OH) complex comes from the propensity of
iron complexes to form Fe−O−Fe dimers. We hypothesized
that the large steric encumbrance of the tris(2,4,6-triphenyl)-
phenyl corrole ligand (ttppcH3) shown in Figure 1 might
prevent μ-oxo dimer formation. Metal-free ttppcH3 was
prepared via modification of a previous report.4a Metalation
Scheme 1. Oxygen Rebound Mechanism for CYP
Received: July 28, 2017
© XXXX American Chemical Society
A
DOI: 10.1021/jacs.7b07979
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Communication
ν(Fe−18OH) mode on the basis of the 25 cm−1 downshift
predicted for a diatomic Fe−O oscillator (Figure S6). The
energy of this vibration falls in the same range as ν(FeIV−OH)
seen for the heme protein chloroperoxidase (ν(FeIV−OH) =
565 cm−1).5 A Badger’s rule analysis for a range of stretching
frequencies normalized by reduced mass of the diatomic group6
can be applied to predict an Fe−OH distance of 1.83 Å for 2,
which compares reasonably well with d(Fe−OH) = 1.857(3) Å
obtained from XRD. The sensitivity of the ν(FeIV−OH) mode
to H/D exchange via exposure of 2 to D2O, which results in a 3
cm−1 upshift, supports the Fe−OH structure of 2.
The synthesis of the FeIII complex 3, the expected product
from the rebound process, was also targeted to obtain a
spectroscopic benchmark. The cyclic voltammogram of 1 shows
two reversible redox couples at E1/2 = +0.46, −0.64 V vs Fc+/Fc
(Figure S7a), and the latter couple can be assigned to the one-
electron reduction of 1 to give FeIII(ttppc). Bulk chemical
reduction of 1 was carried out by addition of Cr(C6H6)2 (E°′ =
−1.15 V vs Fc+/Fc, CH2Cl2) in ethyl acetate/diethyl ether,
giving a new red species with λmax = 415, 575, and 765 nm
(Figure S8a). The red species was purified by chromatography
to yield FeIII(OEt2)2(ttppc) (3·2Et2O) as a red solid in good
yield (81%). The UV−vis and 1H NMR data for 3 are
consistent with other FeIII corroles.4b
Figure 1. Displacement ellipsoid plots (35% probability level) for the
molecular structures of (a) 1 and (b) 2 at 110(2) K. Hydrogen atoms,
except for H(1) in 2, omitted for clarity. Inset: structure of the metal-
free corrole ligand used in this study.
We employed electron paramagnetic resonance (EPR) and
Mossbauer spectroscopies to further characterize the electronic
̈
structures of complexes 1−3. Corroles can behave as
noninnocent ligands similar to porphyrins, complicating the
oxidation state assignments of the central metal. The
assessment of the electronic structures of FeIV(corrole)(L)
complexes has been particularly challenging, with the valence
tautomers FeIV(corrole)(L)/FeIII(corrole+•)(L), or a hybrid
structure in between these canonical descriptions, as
possibilities.7 Complexes 1 and 2 are both EPR-silent (X-
with FeCl2 gave FeIVCl(ttppc) (1) as a brown solid (86%). The
UV−visible and 1H nuclear magnetic resonance (NMR)
spectra of 1 are similar to those of other FeIV(Cl) corroles.4b
The structure for 1 was determined by single crystal X-ray
diffraction (XRD), and the result is shown in Figure 1a. The
Fe−N and Fe−Cl bond lengths (Table S3) are similar to those
for FeIV(Cl)(tpfc) (tpfc = tris(pentafluoro)phenyl corrole).4b
Reaction of 1 with Bu4N+OH− was monitored by UV−vis
spectroscopy. Upon addition of Bu4N+OH−, the spectrum for 1
converted to a new species with a single Soret band at 403 nm
(Figure S1). Monitoring the same reaction by 1H NMR
spectroscopy (Figure S2) revealed that paramagnetic 1 converts
to a new species with a distinct pattern of paramagnetically
shifted peaks. Mass spectrometry revealed a positive ion peak at
m/z 1280.6, corresponding to [Fe(ttppc)(OH)]+ (calcd m/z
1280.41). These data are consistent with formation of a new
iron-hydroxide complex, 2. Metathesis with LiOH in a two-
phase H2O/toluene mixture converted 1 to 2 on a larger scale,
allowing for the isolation of 2 in the organic phase. The
structure of FeIV(OH)(ttppc) (2) was confirmed by XRD
(Figure 1b). The Fe−O distance for the terminal hydroxide
ligand is 1.857(3) Å. Although the hydrogen atom of the OH−
ligand could not be located with certainty in the difference
Fourier maps, the Fe−O distance is too long for an
unprotonated, terminal oxo ligand, and charge balance also
requires protonation, presuming a formal +4 oxidation state for
Fe.
band, 16 K), whereas Mossbauer spectroscopy reveals quadru-
̈
pole doublets with similar parameters (1: δ = 0.18 mm/s, ΔEQ
= 2.86 mm/s; 2: δ = 0.13 mm/s, ΔEQ = 2.21 mm/s) (Figure
S9). The EPR spectrum for 3 shows an intense feature at g =
4.3 (Figure S8b) consistent with an intermediate-spin FeIII (S =
3/2) corrole,4b and the Mossbauer spectrum (Figure S9c)
̈
shows a quadrupole doublet (δ = −0.09 mm s−1, ΔEQ = 4.00
mm s−1) that is consistent with loss of magnetic splitting from
intramolecular interactions, as seen for other FeIII porphyrinoid
compounds.8 As seen for other FeIV corroles, the Mossbauer
̈
and EPR data for 1 and 2 do not provide sufficient information
to make a definitive assignment between FeIV(corrole)(L) or
FeIII(corrole+•)(L) descriptions. However, the data are fully
consistent with 1 and 2 being one redox level above the FeIII
complex 3, which is the key feature required to initiate studies
on the rebound process.10
Complex 2 was examined in reactions with carbon-centered
radicals (R•). Trityl radical (Ph3C•) is relatively stable in
organic solutions, in equilibrium with its dimeric state
(Gomberg’s dimer): 2 Ph3C• ⇌ (Ph3C)2 (∼2% Ph3C• at 23
°C).9 Reaction of FeIV(OH)(ttppc) with excess Gomberg’s
dimer in toluene at 23 °C resulted in the rapid, isosbestic
conversion of FeIV(OH)(ttppc) to FeIII(ttppc) by UV−vis
Characterization of 1 and 2 in toluene at 110 K was carried
out by resonance Raman (RR) spectroscopy. Excitation of 1 at
407 nm (Kr laser) gave rise to a prominent, resonance-
enhanced peak at 343 cm−1, which can be assigned to the Fe−
Cl vibrational mode, and confirmed by 57Fe isotopic
substitution (Figure S5). This peak is not observed in the RR
spectrum of 2, whereas a new peak is detected at 576 cm−1.
Isotopic substitution of 2 via exchange with H218O gives 2-18O,
and a new RR band at 555 cm−1 readily assigned to the
1
(Figure S10). H NMR spectroscopy (Figure S11) and gas
chromatography (GC) (Figure S15a) showed that Ph3COH
(trityl alcohol) was formed in good yield (77%). Isotopic
labeling of 2 was carried out by exchange with excess H218O
under biphasic conditions and confirmed by mass spectrometry.
Reaction of FeIV(18OH)(ttppc) with Ph3C• led to Ph3C18OH
B
DOI: 10.1021/jacs.7b07979
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Communication
with 57% incorporation of 18O (Figure S16). Taken together,
these data show conclusively that OH is transferred from 2 to
the trityl radical in a rebound process.10
Insight into the mechanism of the rebound process was
obtained by examination of a series of trityl derivatives, (p-X-
C6H4)3C• (X = OMe, tBu, Ph, Cl), in which the electron-
donating properties of the para-X substituent was varied
(Figure 2a). These derivatives exist exclusively as monomers in
transition state. For comparison, H atom transfer reactions
between alkoxyl radicals and phenols (RO• + HOAr → ROH
+ •OAr) also show small ρ values, and are described as
proceeding through charge neutral transition states.11
The redox potentials (Eox) for the (p-X-C6H4)3C+/• couples
range from 0.00 to −0.58 V versus Fc+/0 (CH3CN). With the
exception of the −Cl derivative (Eox = 0.00 V), these values are
lower than the redox potential for 2 (−0.11 V versus Fc+/0
)
measured by cyclic voltammetry (Figure S7b), making electron-
transfer (ET) from the radicals to 2 an exergonic process. The
ET process for the −Cl derivative is endergonic by 2.5 kcal/
mol. To determine the involvement of ET in the rebound
mechanism, a Marcus plot of (RT/F) ln(k) vs Eox was
constructed (Figure 2d). The Marcus cross-relation indicates
that a slope of ∼ −0.5 is expected for a simple, rate-limiting ET
process.12 The Marcus plot in Figure 2d reveals a linear
correlation for the rebound reaction, but the slope is small
(−0.15). This slope can be compared to the small slope
(−0.05) seen for the reaction of cumylperoxyl radical and
phenols, which is a good model of a concerted H atom transfer,
as opposed to ET-type, mechanism.13a,b In contrast, Fujii has
reported large slopes (0.60−0.68) for aromatic hydroxylation of
benzene derivatives by FeIV(O)(porph+•) complexes. Simple
outer-sphere ET for these reactions was ruled out based on
endergonic ΔG°ET values, but the large Marcus slopes pointed
to an ET pathway that was coupled to a subsequent favorable
C−O bond formation step.13c
With the data now in hand from the direct examination of a
radical rebound reaction, we can suggest a straightforward
mechanistic paradigm to assess the rebound step, in analogy
with H atom transfer. HAT proceeds in an approximately
concerted fashion, or in more of a stepwise one in which the
ET/PT steps are coupled to varying degrees. The rebound
mechanism can be parsed in a similar model, as shown in
Figure 3a, in which rebound can occur via a concerted pathway,
Figure 2. (a) Oxygen rebound reaction for 2 with para-X-substituted
trityl radicals. (b) Time-resolved UV−vis spectral changes observed
upon addition of (p-tBu-C6H4)3C• (0.33 mM) to 2 (20 μM) (red line,
t = 0.5 s) to form 3 (green line, t = 220 s) in toluene at 23 °C. Inset:
changes in absorbance for 570 nm vs time. (c) Hammett plot and (d)
Marcus plot.
Figure 3. (a) Scheme depicting concerted and separated ET/CT steps
for radical rebound. (b) Charge separation in the transition state of the
rebound reaction.
solution. Initial studies began with the p-tBu derivative.
Reaction of (p-tBu-C6H4)3C• with 2 gave the expected
rebound products, 3 and (p-tBu-C6H4)3COH, in good yields
(Figure S13, S15b). Monitoring this reaction by UV−vis under
pseudo-first-order conditions (Figure 2b) yielded a pseudo-
first-order rate constant (kobs), and a plot of kobs versus [(p-tBu-
C6H4)3C•] was linear, whose slope gave a second-order rate
constant of 49(1) M−1 s−1. Second-order rate constants for the
other substituted trityl radicals were determined similarly,
ranging from 12.6(1) to 357(4) M−1 s−1 for the methoxy
derivative (Table S4).
or electron-transfer and “cation transfer” (CT) steps. The
Hammett and Marcus plots in Figures 2c,d indicate that the
radical rebound reactions for 2 follow a concerted (diagonal
path, Figure 3a), as opposed to “ET/CT” type of mechanism,
in which C−O bond formation occurs concomitant with
reduction of the iron complex. Thus, the transition state of the
concerted rebound reaction is likely closer to the charge neutral
structure shown in Figure 3b. Although 2 is sterically crowded,
examination of a space-filling model of 2 (Figure S23) suggests
that there is space for the trityl radical to combine with the
FeOH above the pyrrole−pyrrole bond.
A plot of log k versus 3σ+para, where k is the second order rate
constant and σ+ is the Hammett parameter for the para-X
para
substituents, is shown in Figure 2c. A linear correlation is
obtained (r2 = 0.99), but the magnitude of the slope (ρ =
−0.55) is small, suggesting only a small charge separation in the
The mechanism of action for heme and nonheme oxy-
genases, as well as related transition metal catalysts, has been
C
DOI: 10.1021/jacs.7b07979
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Communication
342, 825. (c) Green, M. T.; Dawson, J. H.; Gray, H. B. Science 2004,
304, 1653. (d) A neutron diffraction study provided evidence for a
protonated Cpd-II in a His-ligated ascorbate peroxidase. See: Kwon,
H.; Basran, J.; Casadei, C. M.; Fielding, A. J.; Schrader, T. E.;
Ostermann, A.; Devos, J. M.; Aller, P.; Blakeley, M. P.; Moody, P. C.
E.; Raven, E. L. Nat. Commun. 2016, 7, 13445.
(4) (a) Liu, H.-Y.; Yam, F.; Xie, Y.-T.; Li, X.-Y.; Chang, C. K. J. Am.
Chem. Soc. 2009, 131, 12890. (b) Simkhovich, L.; Mahammed, A.;
Goldberg, I.; Gross, Z. Chem. - Eur. J. 2001, 7, 1041.
(5) Stone, K. L.; Behan, R. K.; Green, M. T. Proc. Natl. Acad. Sci. U. S.
A. 2006, 103, 12307.
(6) Spaeth, A. D.; Gagnon, N. L.; Dhar, D.; Yee, G. M.; Tolman, W.
B. J. Am. Chem. Soc. 2017, 139, 4477.
(7) (a) Walker, F. A.; Licoccia, S.; Paolesse, R. J. Inorg. Biochem.
2006, 100, 810. (b) Ye, S.; Tuttle, T.; Bill, E.; Simkhovich, L.; Gross,
Z.; Thiel, W.; Neese, F. Chem. - Eur. J. 2008, 14, 10839.
(8) Leeladee, P.; Jameson, G. N. L.; Siegler, M. A.; Kumar, D.; de
Visser, S. P.; Goldberg, D. P. Inorg. Chem. 2013, 52, 4668.
(9) Trityl radical has been employed to study analogous NR transfer
reactions from Cu- and Fe-N(H)Ar species. See: (a) Jang, E. S.;
described in the framework of the rebound process shown in
Scheme 1 for over 40 years. However, until now, the critical
rebound step could not be observed and studied directly. In this
report, we have described the first direct examination of a
rebound hydroxylation, in which a well-defined Fe(OH)
porphyrinoid complex is reacted with a series of carbon
radicals. A key to this study was the use of a corrole ligand to
isolate an Fe(OH) complex that is at the same oxidation level
as the Cpd-II intermediate in heme enzymes. Such a species
was previously unknown in porphyrinoid models of heme
enzymes. The kinetic measurements provide a mechanistic
assessment of the radical hydroxylation step, leading to a model
for the rebound process that has both concerted and stepwise
pathways, analogous to the well-known paradigm for HAT.
Taken together, the data point to a more concerted process for
rebound. With the precedent established here, the factors that
favor and control rebound hydroxylation over other pathways,
such as desaturation,14a,b halogenation,14c,d or radical cage
escape,14e may now be amenable to direct interrogation.
McMullin, C. L.; Kaß, M.; Meyer, K.; Cundari, T. R.; Warren, T. H. J.
̈
Am. Chem. Soc. 2014, 136, 10930. (b) Iovan, D. A.; Betley, T. A. J. Am.
Chem. Soc. 2016, 138, 1983.
ASSOCIATED CONTENT
* Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/jacs.7b07979.
■
(10) It is possible that the trityl alcohol product could remain
coordinated to the FeIII center, or exist in an equilibrium between
bound and unbound states in solution.
(11) Colclough, N.; Smith, J. R. L. J. Chem. Soc., Perkin Trans. 2 1994,
2, 1139.
S
Experimental details (PDF)
(12) Marcus, R. A.; Sutin, N. Biochim. Biophys. Acta, Rev. Bioenerg.
1985, 811, 265.
Data for C91H59ClFeN4 (CIF)
Data for C91H60FeN4O•C5H12 (CIF)
(13) (a) Osako, T.; Ohkubo, K.; Taki, M.; Tachi, Y.; Fukuzumi, S.;
Itoh, S. J. Am. Chem. Soc. 2003, 125, 11027. (b) Lee, J. Y.; Peterson, R.
L.; Ohkubo, K.; Garcia-Bosch, I.; Himes, R. A.; Woertink, J.; Moore,
C. D.; Solomon, E. I.; Fukuzumi, S.; Karlin, K. D. J. Am. Chem. Soc.
2014, 136, 9925. (c) Asaka, M.; Fujii, H. J. Am. Chem. Soc. 2016, 138,
8048.
AUTHOR INFORMATION
Corresponding Author
*dpg@jhu.edu
ORCID
■
(14) (a) Grant, J. L.; Mitchell, M. E.; Makris, T. M. Proc. Natl. Acad.
Sci. U. S. A. 2016, 113, 10049. (b) Rettie, A. E.; Rettenmeier, A. W.;
Howald, W. N.; Baillie, T. A. Science 1987, 235, 890. (c) Wong, S. D.;
Srnec, M.; Matthews, M. L.; Liu, L. V.; Kwak, Y.; Park, K.; Bell, C. B.,
III; Alp, E. E.; Zhao, J.; Yoda, Y.; Kitao, S.; Seto, M.; Krebs, C.;
Bollinger, J. M.; Solomon, E. I. Nature 2013, 499, 320. (d) Galonic, D.
P.; Barr, E. W.; Walsh, C. T.; Bollinger, J. M.; Krebs, C. Nat. Chem.
Biol. 2007, 3, 113. (e) Cho, K.-B.; Hirao, H.; Shaik, S.; Nam, W. Chem.
Soc. Rev. 2016, 45, 1197.
Jan Paulo T. Zaragoza: 0000-0002-2583-6571
Pierre Moenne-Loccoz: 0000-0002-7684-7617
̈
David P. Goldberg: 0000-0003-4645-1045
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The authors acknowledge support by the NIH (Grant
GM101153 to D.P.G. and Grant GM101390 to M.T.G.).
J.P.T.Z. thanks JHU for the Glen E. Meyer ’39 Fellowship. We
also thank Dr. R. Baglia for helpful discussions, and T. Pangia
for the synthesis of some compounds.
REFERENCES
■
(1) (a) Ortiz de Montellano, P. R. Chem. Rev. 2010, 110, 932.
(b) Kovaleva, E. G.; Lipscomb, J. D. Nat. Chem. Biol. 2008, 4, 186.
(c) Chen, M. S.; White, M. C. Science 2007, 318, 783. (d) Kudrik, E.
́
V.; Afanasiev, P.; Alvarez, L. X.; Dubourdeaux, P.; Clemancey, M.;
Latour, J.-M.; Blondin, G.; Bouchu, D.; Albrieux, F.; Nefedov, S. E.;
Sorokin, A. B. Nat. Chem. 2012, 4, 1024. (e) Rittle, J.; Green, M. T.
Science 2010, 330, 933. (f) Huang, X.; Groves, J. T. J. Biol. Inorg. Chem.
2017, 22, 185. (g) Shaik, S.; Kumar, D.; de Visser, S. P.; Altun, A.;
Thiel, W. Chem. Rev. 2005, 105, 2279.
(2) (a) Boaz, N. C.; Bell, S. R.; Groves, J. T. J. Am. Chem. Soc. 2015,
137, 2875. (b) Hill, E. A.; Weitz, A. C.; Onderko, E.; Romero-Rivera,
A.; Guo, Y.; Swart, M.; Bominaar, E. L.; Green, M. T.; Hendrich, M.
P.; Lacy, D. C.; Borovik, A. S. J. Am. Chem. Soc. 2016, 138, 13143.
(3) (a) Yosca, T. H.; Langston, M. C.; Krest, C. M.; Onderko, E. L.;
Grove, T. L.; Livada, J.; Green, M. T. J. Am. Chem. Soc. 2016, 138,
16016. (b) Yosca, T. H.; Rittle, J.; Krest, C. M.; Onderko, E. L.;
Silakov, A.; Calixto, J. C.; Behan, R. K.; Green, M. T. Science 2013,
D
DOI: 10.1021/jacs.7b07979
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX