An experimental and quantum chemical investigation of CO binding to heme proteins and model systems: A unified model based on 13C, 17O, and 57Fe nuclear magnetic resonance and57Fe mossbauer and infrared spectroscopies

Article abstract of DOI:10.1021/ja973272j

We have investigated the question of how CO ligands bind to iron in metalloporphyrins and metalloproteins by using a combination of nuclear magnetic resonance (NMR), ⁵⁷Fe Mossbauer, and infrared spectroscopic techniques, combined with density functional theoretical calculations to analyze the spectroscopic results. The results of ¹³C NMR isotropic chemical shift, ¹³C NMR chemical shift anisotropy, ¹⁷O NMR isotropic chemical shift, ¹⁷O nuclear quadrupole coupling constant, ⁵⁷Fe NMR isotropic chemical shift, ⁵⁷Fe Mossbauer quadrupolar splitting, and infrared measurements indicate that CO binds to Fe in a close to linear fashion in all conformational substates. The ¹³C-isotropic shift and shift anisotropy for an A(o) substate model compound: Fe(5,10,15,20-tetraphenylporphyrin)(CO)(N-methylimidazole), as well as the ¹⁷O chemical shift, and the ¹⁷O nuclear quadrupole coupling constant (NQCC) are virtually the same as those found in the A(o) substate of Physeter catodon CO myoglobin and lead to most probable ligand tilt (τ) and bend (β) angles of 0°and 1°when using a Bayesian probability or Z surface method for structure determination. The infrared V(co) for the model compound of 1969 cm^-1 is also that found for A(o) proteins. Results for the A₁ substate (including the ⁵⁷Fe NMR chemical shift and Mossbauer quadrupole splitting) are also consistent with close to linear and untilted Fe-C-O geometries (τ = 4°, β = 7°), with the small changes in ligand spectroscopic parameters being attributed to electrostatic field effects. When taken together, the ¹³C shift, ¹³C shift anisotropy, ¹⁷O shift, ¹⁷O NQCC, ⁵⁷Fe shift, ⁵⁷Fe Mossbauer quadrupole splitting, and v(co) all strongly indicate very close to linear and untilted Fe-C-O geometries for all carbonmonoxyheme proteins. These results represent the first detailed quantum chemical analysis of metal-ligand geometries in metalloproteins using up to seven different spectroscopic observables from three types of spectroscopy and suggest a generalized approach to structure determination.

Full text of DOI:10.1021/ja973272j

4784
J. Am. Chem. Soc. 1998, 120, 4784-4797
An Experimental and Quantum Chemical Investigation of CO Binding
to Heme Proteins and Model Systems: A Unified Model Based on
17
57
57
13C, O, and Fe Nuclear Magnetic Resonance and Fe Mo¨ssbauer
and Infrared Spectroscopies
Michael T. McMahon,† Angel C. deDios,‡ Nathalie Godbout, Renzo Salzmann,
David D. Laws,§ Hongbiao Le,| Robert H. Havlin,§ and Eric Oldfield*
Contribution from the Department of Chemistry, UniVersity of Illinois at Urbana-Champaign,
600 South Mathews AVenue, Urbana, Illinois 61801
ReceiVed September 17, 1997
Abstract: We have investigated the question of how CO ligands bind to iron in metalloporphyrins and
metalloproteins by using a combination of nuclear magnetic resonance (NMR), 57Fe Mo¨ssbauer, and infrared
spectroscopic techniques, combined with density functional theoretical calculations to analyze the spectroscopic
results. The results of 13C NMR isotropic chemical shift, 13C NMR chemical shift anisotropy, 17O NMR
isotropic chemical shift, 17O nuclear quadrupole coupling constant, 57Fe NMR isotropic chemical shift, 57Fe
Mo¨ssbauer quadrupolar splitting, and infrared measurements indicate that CO binds to Fe in a close to linear
fashion in all conformational substates. The 13C-isotropic shift and shift anisotropy for an Ao substate model
compound: Fe(5,10,15,20-tetraphenylporphyrin)(CO)(N-methylimidazole), as well as the 17O chemical shift,
and the 17O nuclear quadrupole coupling constant (NQCC) are virtually the same as those found in the Ao
substate of Physeter catodon CO myoglobin and lead to most probable ligand tilt (τ) and bend (â) angles of
0° and 1° when using a Bayesian probability or Z surface method for structure determination. The infrared
νCO for the model compound of 1969 cm-1 is also that found for Ao proteins. Results for the A1 substate
(including the 57Fe NMR chemical shift and Mo¨ssbauer quadrupole splitting) are also consistent with close to
linear and untilted Fe-C-O geometries (τ ) 4°, â ) 7°), with the small changes in ligand spectroscopic
parameters being attributed to electrostatic field effects. When taken together, the 13C shift, 13C shift anisotropy,
17O shift, 17O NQCC, 57Fe shift, 57Fe Mo¨ssbauer quadrupole splitting, and νCO all strongly indicate very close
to linear and untilted Fe-C-O geometries for all carbonmonoxyheme proteins. These results represent the
first detailed quantum chemical analysis of metal-ligand geometries in metalloproteins using up to seven
different spectroscopic observables from three types of spectroscopy and suggest a generalized approach to
structure determination.
Introduction
function.5 Typically, X-ray crystallographic methods have been
used to deduce metal-ligand geometries, but quite large
variabilities appear to exist between different structures. For
example, Fe-C-O bond angles from 176° to 118° have been
reported,6,7 Fe-O-O angles from 159° to 115° have been
described,8,9 Fe-C-N angles of 180° to 96° have been observed
in alkyl isocyanide adducts,10,11 etc., and it is unclear to what
extent these differences are real. In the case of CO, the results
of infrared and Raman spectroscopy have been used to deduce
a close-to-linear and untilted Fe-C-O geometry,12,13 with the
changes in the observables being attributed to electrostatic field
How small ligands such as CO, O2, RNC, RNO, NO, and
-
N3 bind to metalloproteins and metalloporphyrins has been a
topic of interest for some time,1-4 because of the obvious
importance of especially O2 carriers in respiration, and more
recently of NO and CO ligands as regulators of cell and organ
† National Institutes of Health Molecular Biophysics Training Grant
Trainee (grant GM-08276).
‡ Present address: Department of Chemistry, Georgetown University,
37th and O Streets, N.W., Washington, DC 20057-2222.
§ Present address: Department of Chemistry, University of California,
Berkeley, CA 94720.
| Present address: Laboratorium fu¨r Physicalische Chemie, ETH, Uni-
versitatstr. 22, 8092, Zu¨rich, Switzerland.
(5) Suematsu, M.; Wakabayashi, Y.; Ishimura, Y. CardioVas. Res. 1996,
32, 679-686.
(1) Pauling, L.; Coryell, C. D. Proc. Natl. Acad. Sci. U.S.A. 1936, 22,
210-216.
(2) Pauling, L. Nature 1964, 203, 182-183. Weiss, J. J. Nature 1964,
203, 182-183.
(3) St. George, R. C.; Pauling, L. Science 1951, 114, 629-634. Lein,
A.; Pauling, L. Proc. Natl. Acad. Sci. U.S.A. 1956, 42, 51-54.
(4) Filehne, W. Arch. Exptl. Pathol. Pharmacol. 1878, 9, 329-379.
Wa¨rburg, O.; Kubowitz, T.; Christian, W. Biochem. Z. 1931, 242, 170-
205. Jung, F. Naturwissenschaften 1940, 28, 264-265. Keilin, D.; Hartree,
E. F. Nature 1943, 151, 390-391.
(6) Quillin, M. L.; Arduini, R. M.; Olson, J. S.; Phillips, G. N., Jr. J.
Mol. Biol. 1993, 234, 140-155.
(7) Kuriyan, J.; Wilz, S.; Karplus, M.; Petsko, G. A. J. Mol. Biol. 1986,
192, 133-154.
(8) Shaanan, B. J. Mol. Biol. 1983, 171, 31-59. Steigemann, W.; Weber,
E. In Hemoglobin and Oxygen Binding; Ho, C., Ed.; Elsevier Biomedical:
New York, 1982; pp 19-24.
(9) Phillips, S. E. V. J. Mol. Biol. 1980, 142, 531-554.
(10) Mims, M. P.; Olson, J. S.; Russu, I. M.; Miura, S.; Cedel, T. E.;
Ho, C. J. Biol. Chem. 1983, 258, 6125-6134.
S0002-7863(97)03272-1 CCC: $15.00 © 1998 American Chemical Society
Published on Web 05/02/1998

CO Binding to Hemes
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4785
zole) will be described in detail elsewhere27 and basically followed
standard methods.28 The exception is that two samples were prepared.
The first was prepared in a “standard” fashion in that solvent was
removed from crystals in vacuo, while in a second sample the mother
liquor was retained, in an effort to retain good quality crystals for solid-
state NMR. CO (13CO, 99%; Cambridge Isotope Laboratories, An-
dover, MA; C17O, 37%, Icon Services, Inc., Summit, NJ) gas exchange
was achieved prior to crystallization using a Pd/Al2O3 heterogeneous
catalyst.29 Here, approximately 100 mg Fe(TPP)(CO)(NMeIm) (TPP
) 5,10,15,20-tetraphenylporphyrin; NMeIm ) N-methylimidazole) was
exchanged with 13CO or C17O (∼5 mL at STP) in a heptane solvent at
23 °C using 3 mg of Pd (10%) on Al2O3 (Alfa Products, Danvers, MA).
This catalyst causes essentially statistical (random) exchange in ∼5
min and can be readily removed by filtration. Crystals were then grown
from CH2Cl2/pentane using a gradient technique. Crystallization times
varied but were typically ∼5 days.
effects, much as with our own NMR/IR correlations.14-16
However, very recent crystallographic results on carbonmon-
oxymyoglobin again suggest distorted (∼45-55°) Fe-C-O
bonding,17 consistent with earlier crystallographic results7,18 and
with XANES and EXAFS results.19,20 Distortions have also been
suggested from solid-state NMR21 and from some Mo¨ssbauer
work.22,23 Given the fact that IR/Raman experiments are more
difficult with non-CO ligands, and that the actual detailed
assignments themselves are not always simple to make,24 it
therefore seems to be desirable to explore the possibilities of
using a variety of spectroscopic observables to investigate
metal-ligand bonding. In this paper, we use up to six additional
spectroscopic parameters, the isotropic carbon-13 NMR chemi-
cal shift (δi 13C), the 13C NMR chemical shift anisotropy (CSA)
(∆δ 13C), the isotropic oxygen-17 NMR chemical shift (δi 17O),
the 17O nuclear quadrupole coupling constant (NQCC) (e2qQ/
h, 17O), the isotropic iron-57 NMR chemical shift (δi 57Fe), the
iron-57 Mo¨ssbauer quadrupole splitting (∆EQ, 57Fe), and the
CO infrared stretch frequency (νCO), to test various structural
models, using parameter surfaces and the Bayesian probability
method reported previously for amino acids.25,26 The methods
described are general and show how NMR, Mo¨ssbauer, and IR
techniques can all be used together to develop and test models
of metal-ligand interactions in metalloproteins, by using
quantum chemical methods to make the necessary correlations
between the spectroscopic observables and structure.
For protein NMR spectroscopy, we used both horse heart and sperm
whale myoglobins (Sigma Chemical Company, St. Louis, MO). For
13C and 17O NMR, we typically purified myoglobin (Mb) by using a
Sephadex G25 column and then grew needle or star-shaped crystal
clusters from (NH4)2SO4 (83%, pH ) 7.0). The crystals were then
reduced on a Schlenk line using a few drops of (NH4)2 SO4 solution
containing Na2S2O4 (Sigma), and carbonylated with CO gas. Fourier
transform infrared spectra were recorded on a Nicolet (Madison, WI)
Magna-IR 750 spectrometer in order to deduce substate composition.
The Ao substate of MbCO was produced in ∼95% yield by
converting Mb into the tetrazole derivative of His6430,31 with CNBr
(Aldrich Chemical Co., Milwaukee, WI) and NaN3 (Fisher Scientific,
Fair Lawn, NJ). Crystals were grown from (NH4)2SO4, (78%, pH )
6.7) followed by reduction and carbonylation as described above.
Sperm whale MbCO was used for 57Fe NMR and Mo¨ssbauer
spectroscopy, and was from the same batch whose preparation has been
described previously.32
Experimental Section
Chemical Aspects. The synthesis and characterization of the model
compound Fe(5,10,15,20-tetraphenylporphyrin)(CO)(N-methylimida-
(11) Johnson, K. A.; Olson, J. S.; Phillips, G. N., Jr. J. Mol. Biol. 1989,
207, 459-463. Eich, R. F.; Li, T.; Lemon, D. D.; Doherty, D. H.; Curry,
S. R.; Aitken, J. F.; Mathews, A. J.; Johnson, K. A.; Smith, R. D.; Phillips,
G. N. Jr.; Olson, J. S. Biochemistry 1996, 35, 6976-6983. Johnson, K. A.;
Olson, J. S.; Smith, R. D.; Phillips, G. N., Jr.Protein Data Bank in
Crystallographic Databases - Information Content, Software Systems,
Scientific Applications; Data Commission of the International Union of
Crystallography; file nos. 2MYA, 2MYB, 2MYC, and 2MYD, 2MYE,
1TES.
(12) Ray, G. B.; Li, X.-Y.; Ibers, J. A.; Sessler, J. L.; Spiro, T. G. J.
Am. Chem. Soc. 1994, 116, 162-176. Ivanov, D.; Sage, J. T.; Keim, M.;
Powell, J. R.; Asher, S. A.; Champion, P. M. J. Am. Chem. Soc. 1994, 116,
4139-4140.
(13) Lim, M.; Jackson, T. A.; Anfinrud, P. A. Science 1995, 269, 962-
965.
Spectroscopic Aspects. Solid-state 13C and 17O NMR spectra were
recorded using “home-built” spectrometers, which consist of 8.45 T
3.5 in. bore and 11.7 T 2.0 in. bore superconducting solenoid magnets
(Oxford Instruments, Osney Mead, Oxford, U.K.), and Tecmag
(Houston, TX) Aries and Libra data systems, together with a variety
of other digital and rf circuitries. For rf pulse amplification, we used
Amplifier Research (Souderton, PA) and Henry Radio (Los Angeles,
CA) transmitters. Solid-state spectra were obtained by using 5 mm
Doty Scientific (Columbia, SC) “magic-angle” sample-spinning (MAS)
NMR probes, in some cases using sealed Wilmad (Buena, NJ) glass
inserts. The 90° pulse widths varied but were typically ∼4 µs for 13
C
and 5 µs (solid 90° pulse) for 17O. For 13C NMR, we typically used
1H-13C cross polarization33 with nonquaternary suppression.34 Recycle
times for 13C NMR were typically ∼2 s, while for 17O NMR, recycle
times varied from 50 ms (A1 MbCO) to 1 s (Fe(TPP)(CO)(NMeIm)).
Carbon-13 chemical shifts were referenced with respect to the low-
field peak of adamantane, taken to be 38.5 ppm downfield from external
tetramethylsilane, while 17O spectra were referenced with respect to
an external sample of tap water. The convention that high-frequency,
low-field, paramagnetic or deshielded shifts are more positive (Inter-
national Union of Pure and Applied Chemistry, IUPAC, δ scale) was
used. Shift tensor elements use the convention that δ11 g δ22 > δ33.
The principal components of the chemical shift tensor, δ11, δ22, and
δ33, were deduced from the intensities of the individual spinning
(14) Park, K. D.; Guo, K.; Adebodun, F.; Chiu, M. L.; Sligar, S. G.;
Oldfield, E. Biochemistry 1991, 30, 2333-2347.
(15) Augspurger, J. D.; Dykstra, C. E.; Oldfield, E. J. Am. Chem. Soc.
1991, 113, 2447-2451.
(16) Oldfield, E.; Guo, K.; Augspurger, J. D.; Dykstra, C. E. J. Am. Chem.
Soc. 1991, 113, 7537-7541.
(17) Yang, F.; Phillips, G. N., Jr. J. Mol. Biol. 1996, 256, 762-774.
(18) Norvell, J. C.; Nunes, A. C.; Schoenborn, B. P. Science 1975, 190,
568-570.
(19) Ascone, I.; Bianconi, A.; Dartyge, E.; Della Longa, S.; Fontaine,
A.; Momentau, M. Biochim. Biophys. Acta 1987, 915, 168-171.
(20) Bianconi, A.; Congiu-Castellano, A.; Durham, P. J.; Hasnain, S.
S.; Phillips, S. Nature 1985, 318, 685-687.
(21) Gerothanassis, I.; Barrie, P. J.; Momentau, M.; Hawkes, G. E. J.
Am. Chem. Soc. 1994, 116, 11944-11949.
(22) Trautwein, A.; Maeda, Y.; Harris, F. E.; Formanek, H. Theor. Chim.
Acta 1974, 36, 67-76.
(27) Salzmann, R.; Ziegler, C.; Suslick, K. S.; Oldfield, E. Unpublished
results.
(23) Trautwein, A. In Structure and Bonding; Dunitz, J. D., Hemmerich,
P., Holm, R. H., Ibers, J. A., Jørgensen, C. K., Neilands, J. B., Reinen, D.,
Williams, R. J. P., Eds.; Springer-Verlag: New York, 1974; Vol. 20, pp
101-166.
(24) Hirota, S.; Ogura, T.; Shinzawa-Itoh, K.; Yoshikawa, S.; Nagai,
M.; Kitagawa, T. J. Phys. Chem. 1994, 98, 6652-6660. Hu, S.; Vogel, K.
M.; Spiro, T. G. J. Am. Chem. Soc. 1994, 116, 11187-11188. Tsuboi, M.
Ind. J. Pure Appl. Phys. 1988, 26, 188-191.
(25) Le, H.; Pearson, J. G.; deDios, A. C.; Oldfield, E. J. Am. Chem.
Soc. 1995, 117, 3800-3807.
(26) Heller, J.; Laws, D. D.; Tomaselli, M.; King, D. S.; Wemmer, D.
E.; Pines, A.; Havlin, R. H.; Oldfield, E. J. Am. Chem. Soc. 1997, 119,
7827-7831.
(28) Peng, S.-M.; Ibers, J. A. J. Am. Chem. Soc. 1976, 98, 8032-8036.
(29) Webb, A. N.; Mitchell, J. J. J. Phys. Chem. 1959, 63, 1878-1885.
Noack, K.; Ruch, M. J. Organomet. Chem. 1969, 17, 302-322.
(30) Adachi, S.-I.; Morishima, I. Biochemistry 1992, 31, 8613-8618.
(31) Kamiya, N.; Shiro, Y.; Iwata, T.; Iizuka, T.; Iwasaki, H. J. Am.
Chem. Soc. 1991, 113, 1826-1829.
(32) Lee, H. C.; Gard, J. K.; Brown, T. L.; Oldfield, E. J. Am. Chem.
Soc., 1985, 107, 4087-4088. Chung, J.; Lee, H. C.; Oldfield, E. J. Magn.
Reson. 1990, 90, 148-157.
(33) Pines, A.; Gibby, M. G.; Waugh, J. S. J. Chem. Phys. 1973, 59,
569-590.
(34) Opella, S. J.; Frey, M. H. J. Am. Chem. Soc. 1979, 101, 5854-
5855.

4786 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
McMahon et al.
We next evaluated the 13C and 17O shieldings and shielding tensor
elements, this time as a function of τ and â. Here, the larger Fe bis-
(amidinato)(CO)(NMeIm) model shown in Figure 1B was used, since
we were more interested in investigating the actual shifts, rather than
the correlations between shifts and νCO, and the elimination of the full
vibrational analysis made computation of a shielding surface containing
35 τ,â values feasible. The nuclear shieldings were calculated with
the sum-over-states density functional perturbation theory (SOS-DFPT)
approach in its LOC1 approximation44 using individual gauges for
localized orbitals (IGLO)42 as implemented in the deMon program.39
We used the same iron basis as that shown above, while the C,O basis
was reduced to IGLO-II.42 Next, we investigated 57Fe NMR chemical
shifts and the electric field gradients at 17O, and 57Fe (directly related
to the 57Fe Mo¨ssbauer quadrupole splittings, ∆EQ), again as a function
of ligand tilt and bend. Since metal shifts have in the past been rather
difficult to evaluate, we used a much larger fragment and increased
the grid coarseness for our τ,â study. We used a published porphyrin
geometry28 minus the Câ-substituents, together with an axial imidazole
oriented in the same way as reported in the crystal structure of
carbonmonoxymyoglobin.45 We carried out geometry optimizations of
the Fe-C and C-O bond lengths at fixed ligand tilt and bend angles
with the Gaussian 94/DFT program46 using the B3LYP hybrid
exchange-correlation functional,47 Wachters’ (62111111/3311111/3111)
Fe basis set,40,41 a 6-31G* basis on the carbonyl group and the 5 nitrogen
atoms, and a 3-21G basis on the remaining atoms. The nuclear
shielding calculations were carried out in Gaussian 94/DFT using gauge-
including atomic orbitals (GIAO),46 a locally dense basis set scheme,
and the B3LYP functional.47 Wachters’ basis set was used on the iron,
6-311++G(2d) on the CO group and five nitrogen atoms with 6-31G*
and 4-31G*, on the second and third atom shells, respectively.
Figure 1. Fragments used for calculations. (A) Fe-C-O model used
in electrostatic field calculations. Hyphens represent 0.4e point charges
located at 2 Å from the Fe atom (the 5 closest nitrogen positions in a
heme plus axial base). (B) Fe bis(amidinato)(CO)(N-methylimidazole)
model system used for 13C,17O shielding calculations. (C) Full porphine
model used for iron shift and efg calculations, Fe(porphine)(CO)(N-
methylimidazole).
sidebands using several versions of the Herzfeld-Berger method,35 as
described below.
Computational Aspects. We carried out five series of density
functional theory (DFT) calculations. In the first, we investigated how
uniform and nonuniform electric fields influence the vibrational
frequency and the 17O nuclear quadrupole coupling constant, as an
extension of previous and more recent work on 13C NMR/17O NMR/
IR shift correlations.14-16,36 These calculations used the small cluster
shown in Figure 1A, which facilitated a full E-field vibrational/shift/
electric field gradient (efg) analysis in a reasonable time period. The
second set of calculations involved a study of how ligand tilt (τ) and
bend (â) alone affect the 13C and 17O shifts and shift anisotropies. This
time a much larger model was used (Figure 1B), an Fe bis(amidinato)-
(CO)(N-methylimidazole) system, which has been investigated previ-
ously by several groups.37,38 Third, we investigated the 57Fe shifts and
57Fe Mo¨ssbauer quadrupolar splittings (equivalent to a determination
of the electric field gradient tensor at the iron nucleus) using a full
porphyrin, the Fe(porphinato)(CO)(N-methylimidazole) system shown
in Figure 1C. Here, a large system was thought to be essential in order
to adequately describe the 57Fe-ligand interactions. We also carried
out a series of calculations on several chromium carbonyls and model
hemes in which the orientation of the axial base was varied. Finally,
we examined the shielding and the localized molecular orbital contribu-
tions to shielding at a single (linear) geometry.
Heme model ligand shielding calculations were also done with G94/
DFT using the BPW91 functional. For Fe(P)(CO)(pyr) (pyr )
pyridine), the iron was represented with the LANLD2DZ effective core
potential (ECP) with a (42111/2111/311) basis set,48 or with the same
ECP,49 6-311++G(2d) on the CO group, 6-31G* on the nitrogen atoms,
and 3-21G and STO-3G basis sets on the outermost atoms of the
porphyrin ring. For Fe(TPP)(CO)(NMeIm), we used Wachters’ iron
basis set,40,41 6-311++G(2d) on the CO group and the five nitrogen
atoms, 6-31G* on the carbons bonded to the nitrogen atoms and 3-21G*
on the remaining atoms. For the Fe(C2Cap)(CO)(NMeIm), we used
Wachters’ Fe basis set, 6-311++G(2d) on the CO group and 6-31G*
on the rest of the atoms. This approach is basically that used by Bu¨hl
for 57Fe NMR chemical shift calculations in organometallic systems,
(41) Basis sets were obtained from the Extensible Computational
Chemistry Environment Basis Set Database, Version 1.0, as developed and
distributed by the Molecular Science Computing Facility, Environmental
and Molecular Sciences Laboratory, which is part of the Pacific Northwest
Laboratory, P.O. Box 999, Richland, WA 99352, and is funded by the U.S.
Department of Energy under contract DE-AC06-76RLO 1830. Contact
David Feller, Karen Schuchardt, or Don Jones for further information.
(42) Kutzelnigg, W.; Fleischer, U.; Schindler M. In NMR-Basic
Principles and Progress; Springer: Heidelberg, 1990; Vol. 28, p 1965.
(43) Perdew, J. P.; Wang, Y. Phys. ReV. B. 1992, 45, 13244-13249.
(44) Malkin, V. G.; Malkina, O. L.; Casida, M. E.; Salahub, D. R. J.
Am. Chem. Soc. 1994, 116, 5898-5908. Salahub, D. R. In Theoretical and
Computational Chemistry; Politzer, P., Seminario, J. M., Eds.; Elsevier:
New York, 1995.
(45) Abola, E. E.; Bernstein, F. C.; Bryant, S. H.; Koetzle, T. F.; Weng,
J. Protein Data Bank in Crystallographic Databases - Information Content,
Software Systems, Scientific Applications; Data Commission of the Inter-
national Union of Crystallography; file no. 1MBC.
(46) Frisch, M. J.; Trucks, G. W.; Head-Gordon, M.; Gill, P. M. W.;
Wong, M. W.; Foresman, J. B.; Johnson, B. G.; Schlegel, H. B.; Robb, M.
A.; Repiogle, E. S.; Gomperts, R.; Andres, J. L.; Raghavachari, K.; Binkley,
J. S.; Gonzalez, C.; Martin, R. L.; Fox, D. J.; Defrees, D. J.; Baker, J.;
Stewart, J. J. P.; Pople, J. A. Gaussian94/DFT; Gaussian, Inc.: Pittsburgh,
PA, 1994.
The smallest model system is the one recently used to investigate
charge field perturbation of 13C,17O shift-νCO (IR) correlations in
proteins,36 and consists of the octahedral arrangement of atoms and
point charges shown in Figure 1A. The 0.4e point charges ensure a
diamagnetic ground state and electroneutrality and are placed at 2 Å
from the Fe atom. We used the deMon program39 to evaluate the 17
O
efg and νCO, employing Wachters’ all electron basis set for iron
(14s9p5d/8s5p5d)40,41 together with an IGLO-III basis for C and O42
and the Perdew-Wang (PW91) gradient-corrected exchange-correlation
functional.43 The 17O NQCC and νCO values were evaluated in the
presence of uniform electric fields applied along the Fe-C-O axis,
having magnitudes from 0.02 to -0.02 au (1 au field ) 5.14225 ×
109 V/cm), or in the presence of point changes of (e at 2 Å from the
oxygen, along the Fe-C-O axis.
(35) Herzfeld, J.; Berger, A. J. Chem. Phys. 1980, 73, 6021-6030.
(36) deDios, A. C.; Earle, E. J. Phys. Chem. 1997, 101, 8132-8134.
(37) Strich, A.; Veillard, A. Theor. Chim. Acta 1981, 60, 379-383.
(38) Jewsbury, P.; Yamamoto, S.; Minato, T.; Saito, M.; Kitagawa, T.
J. Am. Chem. Soc. 1994, 116, 11586-11587.
(39) deMon program: Salahub, D. R.; Fournier, R.; Mlynarski, P.; Papai,
I.; St-Amant, A.; Ushio, J. In Density Functional Methods in Chemistry;
Labanowski, J., Andzelm, J., Eds.; Springer: New York, 1991. St-Amant,
A.; Salahub, D. R. Chem. Phys. Lett. 1990, 169, 387-392.
(40) Wachters, A. J. H. J. Chem. Phys. 1970, 52, 1033-1036. Wachters,
A. J. H. IBM Technol. Rep. RJ584 1969.
(47) Becke, A. D. J. Chem. Phys. 1993, 98, 5648-5652. Lee, C.; Yang,
W.; Parr, R. G. Phys. ReV. B. 1988, 37, 785-789.
(48) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 270-283. Hay,
P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 284-298. Hay, P. J.; Wadt,
W. R. J. Chem. Phys. 1985, 82, 299-310.
(49) Dolg, M.; Wedig, U.; Stoll, H.; Preuss, H. J. Chem. Phys. 1987,
86, 866-872.

CO Binding to Hemes
Table 1. 13C, 17O NMR Parameters for Fe Bis(amidinato)(CO)(NMeIm) as a Function of Ligand Tilt and Bend
ligand geometry shielding tensor element
22 (ppm)
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4787
tilt (deg)
bend (deg)
σ
11 (ppm)
σ
σ
33 (ppm)
σi (ppm)
|σ22 - σ11| (ppm)
|σ33 - σ11| (ppm)
13C
0
10
20
30
40
50
0
10
20
30
40
0
10
20
30
40
0
10
20
30
40
0
10
20
30
40
0
10
20
30
40
0
0
0
0
0
0
-182.9
-182.8
-180.1
-172.9
-166.1
-156.9
-183.8
-179.9
-170.9
-160.1
-151.9
-183.9
-180.0
-170.8
-151.0
-139.0
-184.4
-178.9
-175.3
-161.6
-128.2
-184.6
-174.9
-179.0
-171.8
-125.7
-184.9
-174.9
-177.8
-171.1
-133.1
-177.0
-170.1
-160.0
-148.9
-169.9
-166.8
-161.1
-132.9
-63.1
263.1
263.2
263.9
253.1
214.9
115.1
264.2
260.1
257.1
246.9
212.1
270.1
260.0
251.2
239.0
206.0
282.2
259.1
244.7
226.4
191.8
290.4
244.1
231.0
199.2
150.3
296.1
244.1
201.2
135.9
6.9
-29.9
-28.8
-25.8
-17.6
-4.8
13
16
19
40
103
204
14
15
10
19
74
15
18
15
3
46
17
24
29
25
19
23
37
43
44
4
446
446
444
426
381
272
448
440
428
407
364
454
440
422
390
345
467
438
420
388
320
475
419
410
371
276
481
419
379
307
140
461
388
299
156
0
47.1
1.8
10
10
10
10
10
20
20
20
20
20
30
30
30
30
30
40
40
40
40
40
50
50
50
50
50
60
60
60
60
-169.8
-164.9
-160.9
-141.1
-77.9
-168.9
-162.0
-155.8
-148.0
-93.0
-167.4
-154.9
-146.3
-136.6
-109.2
-161.6
-137.9
-136.0
-127.8
-121.7
-154.9
-137.9
-128.8
-120.1
-118.1
-145.0
-131.1
-120.0
-121.9
-29.8
-28.2
-24.9
-18.1
-5.9
-27.6
-27.3
-25.1
-20.0
-8.7
-23.3
-24.9
-25.6
-23.9
-15.2
-18.6
-22.9
-28.0
-33.5
-32.4
-14.6
-22.9
-35.1
-51.8
-81.4
-12.7
-27.8
-47.0
-87.9
30
37
49
51
15
32
39
40
27
284.0
217.9
139.0
7.1
10
20
30
17O
0
10
20
30
40
50
0
10
20
30
40
0
10
20
30
40
0
10
20
30
40
0
10
20
30
40
0
10
20
30
40
0
0
0
0
0
0
-346.3
-345.1
-338.0
-335.1
-348.1
-383.9
-357.9
-342.6
-327.9
-325.0
-336.8
-381.7
-354.9
-319.8
-318.1
-330.2
-429.6
-380.3
-318.2
-310.9
-322.9
-489.0
-425.2
-354.9
-306.9
-314.1
-565.7
-487.7
-423.7
-346.7
-300.7
-650.7
-564.3
-512.0
-454.0
-323.3
-318.1
-282.0
-173.1
13.9
-114.9
-328.9
-313.6
-269.9
-167.0
4.2
-329.7
-314.9
-280.8
-179.1
-21.2
-324.2
-311.3
-303.2
-212.9
-60.9
-319.0
-305.2
-302.9
-267.9
-111.1
-309.7
-299.7
-300.7
-301.7
-174.7
-297.7
-295.3
-301.0
-296.0
410.7
412.9
402.0
341.9
180.9
206.1
417.1
410.4
394.1
329.0
155.2
434.3
417.1
393.2
318.9
131.8
465.6
429.7
394.8
312.1
112.1
503.0
444.8
398.1
312.1
143.9
545.3
457.3
400.3
333.3
319.3
570.3
457.7
381.0
366.0
-86.3
-83.4
-72.7
-55.4
-51.1
-97.6
-89.9
-81.9
-67.9
-54.3
-59.1
-92.4
-84.2
-69.1
-59.4
-73.2
-95.9
-87.3
-75.5
-70.6
-90.6
-101.7
-95.2
-86.6
-87.6
-93.8
-110.0
-110.0
-108.0
-105.0
-52.0
-126.0
-134.0
-144.0
-128.0
23
27
56
162
362
269
29
29
58
158
341
52
40
39
139
309
105
69
15
98
262
170
120
52
757
758
740
677
529
590
775
753
722
654
492
816
772
713
637
462
895
810
713
623
435
992
870
753
619
458
1111
945
824
680
620
1221
1022
893
820
0
10
10
10
10
10
20
20
20
20
20
30
30
30
30
30
40
40
40
40
40
50
50
50
50
50
60
60
60
60
39
203
256
188
123
45
126
353
269
211
158
10
20
30

4788 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
McMahon et al.
Table 2. 57Fe NMR Shielding Parameters for FeP(CO)(NMeIm)
backbone torsion angles φ and ψ. In this work, we use up to
seven different observables, as follows, which should give
improved structural predictions:
as a Function of Ligand Tilt and Benda
ligand geometry shielding tensor elements
isotropic isotropic
1. δi 13C ) the isotropic carbon-13 NMR chemical shift;
tilt
(deg)
bend
(deg)
σ11
(ppm)
σ22
(ppm)
σ33
(ppm)
shielding σi shift ∆i
2. ∆δ 13C ) the 13C NMR chemical shift anisotropy, |δ33
δ11|, also known as the tensor breadth or span;
-
(ppm)
(ppm)
0
0
0
20
20
20
40
40
40
0
20
40
0
20
40
0
-12266 -11180 -10961 -11469
-12626 -12032 -11783 -12147
-17295 -16797 -12126 -15405
-12044 -11271 -10999 -11438
-12330 -11921 -11542 -11931
-14371 -13772 -12071 -13404
-13900 -12937 -11236 -12691
-14930 -13947 -11605 -13494
-17521 -16316 -11982 -15273
8547
9225
12483
8516
9009
10482
9769
3. δi 17O ) the isotropic oxygen-17 NMR chemical shift;
4. e2qQ/h 17O ) the oxygen-17 nuclear quadrupole coupling
constant;
5. δi 57Fe ) the isotropic iron-57 NMR chemical shift;
6. ∆EQ 57Fe ) the iron-57 Mo¨ssbauer quadrupolar splitting;
7. νCO ) the Fe-C-O infrared vibrational stretch frequency.
We consider first how the 13C, 17O, and 57Fe NMR shift
parameters vary with structure (tilt, τ, and bend, â), and we
then investigate the accuracy of the experimental 13C tensor
measurements, as well as how these tensors might be expected
to vary as a function of axial base orientation and how individual
localized molecular orbitals contribute to shielding. We then
discuss experimental results on the Ao substate of MbCO,
comparing them with experimental results obtained on a novel
metalloporphyrin, Fe(5,10,15,20-tetraphenylporphyrin)(CO)(N-
methylimidazole), which has a linear and untilted Fe-C-O
structure.27 We then use a Bayesian probability or Z surface
method25,26 to predict τ and â in the Ao substate of MbCO,
followed by a related analysis of τ,â in the A1 substate, where
electrostatic field effects also influence νCO and the ligand NMR
parameters. Finally, we compare the results of calculations of
the electrostatic field perturbations with experiment and con-
clude with a general model for the three conformational
substates, A0, A1, and A3, in terms of τ,â and electrostatics,
which is consistent with all of the experimental results, an
approach which should be applicable to other ligands as well.
Effects of Ligand Tilt and Bend on NMR Observables.
We first consider the effects of ligand tilt and bend on the
following spectroscopic observables: the 13C isotropic shift, the
13C shift anisotropy, the 17O isotropic shift, and the 57Fe isotropic
shift. Computed results for the model systems described above
are given in Tables 1 and 2.
To begin with, we first compare these results with the
experimental data published previously by ourselves and by
others.14,21,54 In these previous studies, the 13C and 17O NMR
observables for CO in picket fence porphyrin were reported14,21
in which CO is linear and untilted.55 For 13C, at τ ) â ) 0° the
calculated values (converted from the absolute shieldings using
a value of 186 ppm for the absolute shielding of tetramethyl-
silane, ref 56) are δiso ) 215 ppm, |δ22 - δ11| ) 13 ppm, and
|δ33 - δ11| ) 446 ppm, to be compared with experimental values
for the linear CO in picket fence porphyrin of δiso ) 205 ppm,
|δ22 - δ11| ) 8 ppm, and |δ33 - δ11| ) 457 ppm. For 17O, the
calculated values (converted from the absolute shieldings using
a value of 306 ppm for the absolute shielding of liquid water,
20
40
10572
12351
a Calculations were performed in Gaussian-94 using the B3LYP
hybrid exchange correlation functional.
as described elsewhere.50 The chromium carbonyl compounds were
investigated with SOS-DFPT using the chromium DZVP (63321/5211/
41+) basis set found in deMon,39 IGLO-II basis sets on the remaining
atoms, and the PW91 functional. The localized molecular orbital
(LMO) contributions to the shielding in the Fe(C2Cap)(CO)(NMeIm)
model were calculated with SOS-DFPT using the iron DZVP (63321/
5211/41+) basis set found in deMon,39 IGLO-II basis sets on the other
atoms, and the PW91 functional.
Small molecule calculations were carried out on International
Business Machines (Austin, TX) RS/6000 computers (models 340, 350,
360, 365 and 3CT), while the larger systems were investigated using
Silicon Graphics/Cray Research (Mountain View, CA) Origin 200,
Origin-2000, and Power Challenge multiple processor machines. The
geometry optimizations and 57Fe shift and efg calculations were
performed using the Silicon Graphics Power Challenge Array at the
National Center for Supercomputing Applications, located in Urbana,
IL, using up to 16 processors.
Results and Discussion
The rapid improvements in computer speed over the past few
years, together with the increasingly widespread applications
of density functional theory, strongly suggest that it should now
be possible to compute a wide range of spectroscopic observ-
ables in metalloproteins and related model systems.51 The
problem which exists, of course, is that the actual protein
structures of interest are themselves the focus of debate. What
is needed, therefore, are model compounds which have the same
spectroscopic observables as do the proteins, but whose
structures are more certain. Theoretical methods can then be
used to predict the measured experimental parameters, validating
the methods, after which aspects of protein structure can begin
to be predicted using numerous observables measured on
different proteins.
This is exactly the same approach which we have employed
with amino acid residues in proteins in which quantum chemical
methods were used to first reproduce chemical shifts (and
shielding tensors) in amino acids52 and proteins;53 the process
was then “inverted” and structural information derived from the
experimental results in both peptides26 and proteins.25 In work
with peptides, the rms (root-mean-square) deviations between
X-ray experiment and NMR prediction were ∼11° for the
ref 57) are δiso ) 392 ppm, |δ22 - δ11| ) 23 ppm, and |δ33
-
δ11| ) 757 ppm, to be compared with experimental values of
δiso ) 372 ppm, |δ22 - δ11| < 30 ppm, and |δ33 - δ11| ) 814
ppm.14 The errors seen on the isotropic shift are those which
are typically found when a range of organometallic compounds
(50) Bu¨hl, M. Chem. Phys. Lett. 1997, 267, 251-257.
(51) Strohmein, M.; Orendt, A. M.; Facelli, J. C.; Solum, M. S.; Pugmire,
R. J.; Parry, R. W.; Grant, D. M. J. Am. Chem. Soc. 1997, 119, 7114-
7120. Yang, W.; Levy, M. Symposium on Density Functional Theory, A
Satellite Symposium of the 9th International Congress of Quantum
Chemistry; Durham, North Carolina, 1997.
(52) deDios, A. C.; Laws, D. D.; Oldfield, E. J. Am. Chem. Soc. 1994,
116, 7784-7786.
(53) deDios, A. C.; Pearson, J. G.; Oldfield, E. Science 1993, 260, 1491-
1496.
(54) Gerothanassis, I. P.; Momentau, M.; Barrie, P. J.; Kalodimos, C.
G.; Hawkes, G. E. Inorg. Chem. 1996, 35, 2674-2679. Gerothanassis, I.
P.; Momentau, M.; Hawkes, G. E.; Barrie, P. J. J. Am. Chem. Soc. 1993,
115, 9796-9797.
(55) Hoard, J. L. In Porphyrins and Metalloporphyrins; Smith, K. M.,
Ed.; Elsevier: New York, 1975; pp 317-380.
(56) Jameson, K. J.; Jameson, C. J. Chem. Phys. Lett. 1987, 134, 461-
466.
(57) Wasylishen, R. E.; Mooibroek, S.; Macdonald, J. B. J. Chem. Phys.
1984, 81, 1057-1059.

CO Binding to Hemes
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4789
Table 3. Computed 13C and 17O NMR Shielding Results for Chromium Carbonyls and Model Hemes
system
σ
11 (ppm)
σ
22 (ppm)
13C
σ
33 (ppm)
σi (ppm)
|σ11 - σ22| (ppm)
Cr(CO)5(pyr), 0° a
-155.3
-155.4
-153.0
-153.4
-188.9
-188.6
-173.6
-199.2
-196.6
-198.9
-198.3
-154.2
-154.4
-151.5
-152.1
-185.6
-186.1
-170.7
-196.5
-194.1
-195.0
-195.8
258.5
258.2
257.3
257.6
301.0
298.8
308.2
308.2
312.9
315.9
314.6
-17.0
-17.2
-15.7
-15.9
-24.5
-25.3
-12.0
-29.2
-25.9
-26.0
-26.5
1.1
1.0
1.5
1.3
3.3
2.5
2.9
2.7
2.5
3.9
2.5
Cr(CO)5(pyr), 45° b
Cr(CO)5(NMeIm), 0° c
Cr(CO)5(NMeIm), 45° d
Fe(TPP)(CO)(pyr)e
Fe(TPP)(CO)(pyr)f
Fe(TPP)(CO)(NMeIm)g
Fe(TPP)(CO)(NMeIm)h
Fe(C2Cap)(CO)(NMeIm)i
Fe(C2Cap)(CO)(NMeIm), 0° j
Fe(C2Cap)(CO)(NMeIm), 45° k
17O
-285.1
-283.2
-267.7
-269.0
-281.8
-283.5
-264.6
-331.6
-316.2
-319.0
-318.4
Cr(CO)5(pyr), 0° a
-286.2
-285.9
-270.8
-272.4
-284.5
-287.0
-265.7
-333.1
-317.6
-320.5
-320.3
355.2
355.5
350.4
369.6
421.1
419.1
449.2
446.5
447.5
451.1
449.0
-72.0
-71.2
-62.7
-63.0
-48.4
-50.4
-27.0
-72.7
-62.1
-62.8
-63.2
1.1
2.7
3.1
3.4
2.7
3.5
1.1
1.5
1.4
1.5
1.9
Cr(CO)5(pyr), 45° b
Cr(CO)5(NMeIm), 0° c
Cr(CO)5(NMeIm), 45° d
FeP(CO)(pyr)e
FeP(CO)(pyr)f
Fe(TPP)(CO)(NMeIm)g
Fe(TPP)(CO)(NMeIm)h
Fe(C2Cap)(CO)(NMeIm)i
Fe(C2Cap)(CO)(NMeIm), 0° j
Fe(C2Cap)(CO)(NMeIm), 45° k
a Structure from ref 75, pyr plane along Cr-Ceq axis, DZVP Cr/IGLO-II basis deMon calculation. b Structure from ref 75, pyr plane 45° to
Cr-Ceq axis, DZVP Cr/IGLO-II basis deMon calculation. c Structure from ref 75 with methylimidazole replacing pyridine, DZVP Cr/IGLO-II
basis deMon calculation, NMeIm plane along Cr-Ceq axis. d Structure from ref 75 with methylimidazole replacing pyridine, DZVP Cr/IGLO-II
basis deMon calculation, NMeIm plane 45° to Cr-Ceq. e Structure from ref 28 (minus the Câ-substituents), Fe LANLD2DZ ECP/(42111/2111/311)
basis set G94/DFT BPW91. See the text for details. f Structure from ref 28 (minus the Câ-substituents), Fe ECP/(42111/2111/311) basis set G94/
DFT BPW91. g Structure from ref 27 (with the Cmeso phenyl groups), riding model refinement, Wachters’ Fe basis set, G94/DFT BPW91. h Structure
from ref 27 (with the Cmeso phenyl groups), Fe-C/C-O geometry optimized, Wachters’ Fe basis set, G94/DFT BPW91. i Structure from ref 60
(minus the ring substituents), Wachters’ basis set, G94/DFT BPW91. j Same as i but with an axial base orientation of 0°. k Same as i but with an
axial base orientation of 45°.
Table 4. Localized Molecular Orbital Contributions to CO
(∼80 ppm) anisotropies, |δ22 - δ11|, seen in hemoglobin,
Shielding in Fe(porphine)(CO)(N-methylimidazole)a
myoglobin, and some model systems21,54,61 imply tilt angles
between 30° and 40°. This translates to an 18 ppm difference
in isotropic chemical shift, while the experimental difference
between the protein and the picket fence porphyrin (linear)
model is only 0.5 ppm. The results of Table 1 therefore indicate
the following: (1) the calculations, structures, and/or theory have
inadequacies, (2) some aspects of the shielding tensor measure-
ments may be inaccurate, and/or (3) τ,â effects are unimportant
in controlling the overall changes in 13C,17O shifts seen in
different proteins.
nucleus
13C
LMO
1s(C)
LP(O)
σ11
σ22
σ33
σi
∆σ
200.0
-53.4 -53.3
-88.6 -88.5
200.0 200.0
200.0
0
56.7
3.3 -34.5
3 × Bd(CtO)
39.9 -45.7 128.5
3 × Bd(Fe-C) -233.5 -230.1
77.5 -128.7 311.0
ΣAO(Fe)
2.6
2.3 -33.6
-9.5 -36.2
Σb
-172.9 -169.6 287.1 -18.4 460.0
-168.9 -166.1 308.4 -8.9 477.3
totalc
17O
1s(O)
LP(O)
270.1 270.1 270.1 270.1
-158.3 -157.8
0
The results obtained for the 57Fe NMR chemical shifts are
rather similar to those we found for the 13C,17O shift results
given in Table 1 and are presented in Table 2. Experimentally,
the isotropic chemical shift for 57Fe in MbCO has been reported
by us and by others to be at ∼8227 ppm downfield from Fe-
(CO)5.32,62 The DFT shielding calculations indicate an ∼8547
ppm deshielding for τ ) â ) 0° (Table 2), but other τ,â values
are also consistent with experiment. However, highly bent
structures such as τ ) 0, â ) 40° are some 4000 ppm too
deshielded and are quite inconsistent with experiment. When
taken together, the NMR results tend to support a linear
geometry; therefore, we next investigate in more detail some
of the NMR observations which tend to support nonlinearity.
The Question of |δ22 - δ11|. One of the more puzzling
results which emerges from the above calculations is that the
large |δ22 - δ11| values, which have been seen experimentally
34.7 -93.8 193.0
93.3 -141.7 352.8
30.3 -93.9 187.8
3 × Bd(CtO) -259.5 -258.7
3 × Bd(Fe-C) -157.5 -154.7
ΣAO(Fe)
0.4
0.3 -7.0
-2.1 -7.4
Σb
-304.8 -300.8 421.4 -61.4 726.2
-299.0 -295.7 425.4 -56.4 724.4
Totalc
a Values given are in ppm and were determined with the deMon
program. b Sum of 5 LMO contributions shown. c Total LMO contribu-
tions.
are investigated, since the absolute isotropic shieldings are
typically in error by at least 10-20 ppm,58,59 and experimental
errors on the individual tensor elements are also typically this
large.60 Thus, for these systems, there is generally good accord
between theory and experiment. However, the known anticor-
relation between 13C and 17O shifts seen in numerous proteins14
is not apparent in the calculated results. In addition, the large
(61) Barrie, P. J.; Gerothanassis, I. P.; Momentau, M.; Hawkes, G. E. J.
Magn. Reson. Ser. B 1995, 108, 185-188. Barrie, P. J.; Gerothanassis, I.
P.; Momentau, M.; Hawkes, G. E. J. Magn. Reson. Ser. B 1995, 108, 185-
188.
(62) LaMar, G. N.; Dellinger, C. M.; Sankar, S. S. Biochem. Biophys.
Res. Commun. 1985, 128, 628-633. Baltzer, L.; Becker, E. D.; Tschudin,
R. G.; Gansow, O. A. J. Chem. Soc., Chem. Commun. 1985, 1040-1041.
(58) Kaupp, M.; Malkin, V. G.; Malkina, O. L.; Salahub, D. R. Chem.
Eur. J. 1996, 2, 24-30.
(59) Salzmann, R.; Kaupp, M.; McMahon, M.; Oldfield, E. J. Am. Chem.
Soc. 1998, 120, 4771-4783.
(60) Hawkes, G. E.; Sales, K. D.; Lian, L. Y.; Gobetto, R. Proc. R. Soc.
London A 1989, 424, 93-111.

4790 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
McMahon et al.
Table 5. Experimental Determinations of Chemical Shift Tensor Elements in Mo(CO)6 and Fe(TPP)(CO)(NMeIm)
system
Mo(CO)6
method
µ
F
δ
11 (ppm)
δ
22 (ppm)
δ
33 (ppm)
δi (ppm)
|δ11 - δ22| (ppm)
static
338
346
350
336
338
328
329
336
-72
-71
-76
-70
201
201
201
201
0
18
21
0
least squaresa
Masfitb
11.71
11.93
11.4
-0.91
-0.90
-1.0
Z surfacec
Fe(TPP)(CO)(NMeIm)
dry
solvated
Z surfacec
Z surfacec
390
370
315
342
-90
-98
205
205
75
28
a A least-squares fitting program kindly provided by R. E. Wasylishen. b A MAS NMR fitting program kindly provided by Alex Pines. c A
Bayesian probability fitting program kindly provided by Dr. H. Le.
in several heme proteins and model compounds21,61 and have
in proteins been attributed to geometric (τ,â) distortions,21 do
not appear in the calculations. Or rather, when they do appear,
they are accompanied by very large (J15 ppm) changes in
isotropic shift from the linear, model compounds, effects which
are not seen experimentally.
We therefore sought to see if such effects might be observ-
able, computationally, in a series of model systems, by rotating
the axial bases trans to CO, to see if, for example, metal dπ f
ligand π* “back-bonding” effects might in some cases cause
anisotropies in the 13C shielding tensor elements, a “proximal
side” effect.38 Results for Cr(CO)5(pyridine) and Cr(CO)5(N-
methylimidazole) are presented in Table 3. Quite clearly, the
anisotropies |σ22 - σ11| due to two different orientations (0°,
45°) of two ligands (pyridine, N-methylimidazole) cause at most
a 1.5 ppm asymmetry of the shielding tensor of the trans-CO,
although the effect is somewhat larger for the cis-COs (data
not shown) and for the 17O shielding tensor elements as well,
Table 3.
We next reasoned that the Cr(CO)5(B) model might simply
be too crude a model to reproduce the effects seen in proteins,
since it lacked the porphyrin macrocycle. We therefore
investigated (a) a linear and untilted Fe-C-O system, FeP-
(CO)(pyr), using two different Fe ECP/basis sets, resulting in a
maximal |σ22 - σ11| of 3.3 ppm, and (b) two Fe(TPP)(CO)-
(NMeIm) structures, one with an exact X-ray structure deduced
using a riding model refinement63 for the terminal CO, the
second containing a DFT/G94 optimized Fe-C-O geometry,
Table 3. The Fe P (pyr) structure contained a relatively planar
porphyrin,64 while that in the N-methylimidazole model is
clearly saddledsbut no effects on |σ22 - σ11| were seen.
Finally, we investigated a structure based on Fe(C2Cap)(CO)-
(NMeIm), an alternative close-to-planar porphyrin, using dif-
ferent axial base geometries, finding again only a 3.9 ppm
maximal value for |σ22 - σ11|, Table 3.
However, the ∆σ 13C result of 477.3 ppm is clearly closer to
experiment, confirming previous observations on metal-olefin
complexes which indicated somewhat more accurate ligand
shieldings using the SOS-DFPT/IGLO approach.
The results of Table 4 are also of interest since they give a
very simple breakdown of the localized molecular orbital (LMO)
contributions to 13C,17O shielding. Apart from the 1s and lone
pair contributions, both 13C and 17O shielding are dominated
by C-O and Fe-C LMO contributions, with Fe-Nim and Fe-
Np contributions being quite minor. This, then, gives a logical
explanation for the lack of major distal-base-orientation effects
on the shieldings and emphasizes once again the unusual nature
of observations of large |σ22 - σ11| values which have been
reported in some systems. Such effects seem not to be possible
on the basis of tilt-bend, since they would be accompanied by
very large changes in isotropic shifts, and are also not due to
proximal side effects.
We then began to investigate to what extent |σ22 - σ11| can
actually be estimated using MAS NMR methods. As a first
test, we reinvestigated the 13C shielding tensors in Mo(CO)6.
In previous work, we reported the spin-echo 13C NMR
spectrum of a static sample of Mo(13CO)6, where it is quite clear
that the shielding tensor is, as expected, axially symmetric with
δ11 ) δ22 ) 338 ppm and δ33 ) -72 ppm, Table 4. However,
two different MAS/SSB fitting programs indicated an ap-
preciable asymmetry, |σ22 - σ11| ) 18 and 21 ppm, Table 5,
using the MAS NMR sideband pattern reported previously.65
In contrast, a Bayesian probability based Z surface method gives
δ11 ) δ22 ) 336 ppm and δ33 ) -72 ppm, Table 5, in good
agreement with the static line shape simulation. The most
probable reason for this observation is, we believe, that any
errors in the fitting procedure, due for example to limited signal-
to-noise ratios, for an axially symmetric tensor will result in
less than axial symmetry, in this case |σ22 - σ11| ≈20 ppm.
Apparently, the Bayesian probability approach gives more
accurate results in such situations, as demonstrated from the
results presented in Table 5.
A second possible source of experimental error might
originate from the sample itself. For example, in a dry sample
there could be heterogeneity, due to crystallographic defects.
We therefore investigated two samples of Fe(TPP)(CO)-
(NMeIm), Figure 2. In Figure 2A, we show the 13C MAS NMR
spectrum of a sample of Fe(TPP)(CO)(NMeIm) which had been
dried in vacuo, while in Figure 2B we show a sample of crystals
in excess mother liquor. The anisotropy seen in the dry sample
is very large, |σ22 - σ11| ) 75 ppm, similar to that reported in
some protein and metalloporphyrin samples,21,61 whereas in the
solvated sample the anisotropy is considerably reduced to ∼28
ppm, Figure 2, a value which is within a 10-15 ppm estimated
uncertainty per tensor element of axially symmetric. Interest-
ingly, there is essentially no change in the isotropic shift, which
is at ∼205 ppm in both cases, Table 4. The anisotropy of the
two tensors, |σ33-σ11|, are very similar also, at 480 and 468 ppm.
Interestingly, the anisotropies deduced using the Gaussian 94
program with the BPW91 functional tend to overestimate the
∆σ values seen experimentally, an effect which we have also
noted in metal-olefin calculations.65 We therefore investigated
the 13C and 17O shieldings again, this time with a full planar
porphyrin and a linear and untilted Fe-C-O fragment, using
SOS-DFPT with the PW91 functional. Here, we again wished
to see to what extent addition of a symmetry-breaking axial
N-methylimidazole might be responsible for δ11 * δ22. Results,
together with the localized molecular orbital contributions to
shielding, are given in Tables 3 and 4. For 13C, |σ22 - σ11| )
3.6 ppm and, for 17O, |σ22 - σ11| ) 3.3 ppm, essentially the
same results obtained with the Gaussian-94 calculations.
(63) Schomaker, V.; Trueblood, K. N. Acta Crystallogr. B. 1968, 24,
63-76.
(64) Kim, K.; Ibers, J. A. J. Am. Chem. Soc. 1991, 113, 6077-6081.
(65) Walter, T. H.; Thompson, A.; Keniry, M.; Shinoda, S.; Brown, T.
L.; Gutowsky, H. S.; Oldfield, E. J. Am. Chem. Soc., 1988, 110, 1065-
1068.

CO Binding to Hemes
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4791
Figure 3. Proton-decoupled (90.46 MHz, 8.45 T) 13C MAS NMR
spectrum of HisE7 tetrazoyl 13CO-myoglobin crystals obtained with
1H cross polarization and nonquaternary suppression at a spinning speed
of 3.4 kHz. Experimental conditions were -25 °C, 71 146 scans, 2 s
recycle time, 3.9 µs (90°) pulse width, 3 ms mix time, 60 µs dephasing
period (NQS), and 100 Hz line broadening due to exponential
multiplication.
Figure 2. Experimental 90.5 MHz 13C CP-MAS NMR spectra of the
A0 heme model, Fe(TPP)(CO)(NMeIm). (A) Dry powder sample
spectrum, 6 µs 90° pulse widths, 5 ms mix time, ωr ) 3.462 kHz. (B)
Solvated crystals, as in A but with ωr ) 3.1 kHz.
Therefore, it appears that while the isotropic shift and shift
anisotropy can be quite reliably evaluated from MAS NMR
spectra when using the Bayesian approach, |δ22 - δ11| values
are clearly more difficult to estimate reliably and perhaps will
be less useful in structural analysis for this reason, especially
when used alone.
The Ao Substate. We next move on to consider just what
information can be deduced about metal-ligand geometries in
proteins by using NMR, Mo¨ssbauer, and IR spectroscopic
results. We first consider the case of the Ao substate of
myoglobin, or heme proteins in general. Here, it is important
to make a connection between the results obtained with relatively
well-characterized model compounds, and with data obtained
on proteins.
Fortunately, the spectroscopic parameters for Fe(TPP)(CO)-
(NMeIm) are very close to those found in the Ao substate of
MbCO. The 13C NMR spectrum of an Ao substate (tetrazole)
MbCO sample is shown in Figure 3, from which we find that
δi 13C ) 205.8 ppm and |σ33-σ11| ) 435 ppm, values which are
within experimental error the same as those found for Fe(TPP)-
(CO)(NMeIm), δi ) 205 ppm and |δ33 - δ11| ) 453 ppm, Table
5. In addition, the 17O-isotropic shifts for Ao proteins and the
TPP model are very similar: δi 17O ) 372 ppm in the TPP
model (exactly the same as found in solid C17O picket fence
porphyrin, ref 14) versus 372 ( 1 ppm for a variety of Ao
substate heme proteins.14 The 17O e2qQ/h value in the TPP
model from a nutation experiment, Figure 4, is 1.0 MHz, to be
compared with 1.1 MHz on average for a range of Ao proteins
from solution T1 measurements,14 essentially the same as the
1.2 MHz we reported previously for C17O picket fence porphyrin
from a line shape simulation.14 In addition, the νCO in the TPP
model, 1969 cm-1 (data not shown) is also that found for Ao
substate proteins.14 Finally, the 57Fe Mo¨ssbauer quadrupole
splitting of 0.35 mm sec-1 in the TPP model66 is close to that
found in Ao substate mutant proteins (P. G. Debrunner, private
communication). Taken together, these results tend to suggest
that the linear Fe-C-O bonded Fe(TPP)(CO)(NMeIm) system
Figure 4. 17O NMR nutation plot for Fe(TPP)(CO)(NMeIm). The
solution 90° pulse width (on H2O) was 10.5 µs. The nutation simulation
yields e2qQ/h ) 1.0 ( 0.2 MHz, with an assumed asymmetry parameter
η ) 0.
and CO picket fence porphyrin, are good model systems for
the Ao substates of heme proteins, since the 13C δi, 13C |σ33-
σ11|, 17O δi, 17O e2qQ/h, νCO, and 57Fe ∆EQ results are very
similar. However, the possibility exists that there are other,
alternative solutions for τ,â, which can also explain all of the
spectroscopic observables, which we next investigate by using
the Bayesian or Z surface approach.25,26
The basic idea here is that if a parameter surface is computed
as a function of tilt and bend, P(τ,â), then the likelihood that a
given experimental value of a parameter, Pexpt, is given by a
particular set of τ,â values is given by25,26
expt - Pcalcdτ,â)2/W
1Z ) e-(P
(1)
where W is a search width parameter.25,26 If the computed value
of the parameter, Pcalcd is identical to that found experimen-
τ,â
tally, Pexpt - Pcalcdτ,â ) 0, then Z ) 1, a high probability solution.
On the other hand, if Pexpt - Pcalcd * 0, then Z f 0, a low
τ,â
probability solution. The method can be readily extended from
a one-parameter or 1Z solution to multiple parameters, so long
(66) Havlin, R. H.; Godbout, N.; Salzmann, R.; Wojdelski, M.; Arnold,
W.; Schulz, C. E.; Oldfield, E., J. Am. Chem. Soc. 1998, 120, 3144-3151.

4792 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
McMahon et al.
Figure 5. NMR parameter surfaces as a function of tilt and bend, computed using the deMon program with the Fe bis(amidinato)(CO)(N-
methylimidazole) model system. (A) 13C-isotropic shielding surface, scaled to δi ) 205 ppm at τ ) â ) 0, the experimental shift value for Fe-
(TPP)(CO)(NMeIm) and picket fence porphyrin. (B) 17O-isotropic shielding surface, scaled to δi ) 372 ppm at τ ) â ) 0. (C) 13C ∆σ (|σ33 - σ11|)
shielding anisotropy surface (ppm). (D) 17O NQCC surface (in MHz).
as the form of P(τ,â) is known. In this case, we investigate
the use of four parameters: the 13C NMR isotropic chemical
shift, the 13C NMR chemical shift anistropy, the 17O NMR
isotropic chemical shift, and the 17O nuclear quadrupole
coupling. The four Z surfaces are then combined as follows:
particular, there are several published results on the 57Fe
Mo¨ssbauer quadrupole splitting, ∆EQ,67 as well as additional
reports on the 57Fe NMR isotropic chemical shift.32,62 We show
in Figure 7 the 13C and 17O MAS NMR spectra of the A1
substate of MbCO, from which we deduce δi, and for 13C, |δ33
- δ11| and these and other parameter results are collected in
Table 6.
4Z(τ,â) )
There are clearly isotropic chemical shift differences, in
addition to 17O quadrupole coupling constant differences,
between the A0 and A1 protein and model system results. These
differences could, in principle, have two origins: either they
are structural or they are electrostatic. In previous work, we
and others have suggested an electrostatic origin for the observed
frequency shifts and NQCC changes in the ligand atoms. The
principal reason for this was that extremely good correlations
are found between the various experimental parameters: δi(17O),
δi(13C), e2qQ/h(17O), and νCO (14) (as shown for example in
Figure 8). Here, we plot three correlations based on protein
NMR/IR results, and in addition, we show the location of the
new model compound, Fe(TPP)(CO)(NMeIm), ∆, as well as
recent δi 13C/νCO results for a C2Cap porphyrin, which has highly
perturbed frequency shifts.68 Essentially all of the protein and
model compound results lie on the same curves, implying a
similar mechanism is responsible for these correlations. Indeed,
with the inclusion of the C2Cap results and results for the
peroxidases, the correlation covers a range of almost 100 cm-1
in νCO, Figure 8.
1Z(13C, δi)1Z(13C, |δ33 - δ11|)1Z(17O, δi)1Z(17O,e2qQ/h) (2)
The actual computed parameter (P) surfaces are shown in
Figure 5 and the individual 1Z surfaces are shown in Figure 6.
Here, it is important to note that the isotropic chemical shift
surfaces have been converted from the chemical shifts computed
theoretically using corrections of 10.3 ppm (13C) and 22.3 ppm
(17O), such that the two known linear and untilted Fe(porphyrin)-
(CO)(NMeIm) models have the experimental isotropic chemical
shift. For the |δ33 - δ11| and 17O NQCC, this conversion was
not necessary. In some Z surfaces, there are high probability
solutions over wide regions of τ,â space, Figure 6. However,
when all four parameters are simultaneously included, only the
τ ) â e 2° region has a high probability, as shown in Figure
6. From this we conclude that the Ao substate of MbCO, and
other heme proteins, has a linear Fe-C-O geometry. This is
certainly not an unreasonable conclusion since all of the
spectroscopic observables are virtually identical between Ao
MbCO and the new Fe(TPP)(CO)(NMeIm) model system. That
is, a linear and untilted Fe-CO is one possible solution.
However, the results of Figures 5 and 6, when considered
together, are very important since they give no evidence for
alternate τ,â conformational solutions which might also fit the
experimental 13C and 17O NMR results.
In earlier work, we reproduced to a modest extent the slopes
of these correlations by using an electrostatic field to polarize
a CO molecule: a model for MbCO. More recently, this model
has been improved to include the electroneutral FeCO species
(67) Debrunner, P. G. In Iron Porphyrins; Lever, A. B. P., Gray, H. B.,
Eds.; VCH Publishers: New York, 1989; Vol. 3, pp 139-234.
(68) Jones, R. D.; Budge, J. R.; Ellis, P. E., Jr.; Linard, J. E.; Summerville,
D. A.; Basolo, F. J. Organomet. Chem. 1979, 181, 151-158.
The A1 Substate. We next consider the A1 substate of
MbCO and other heme proteins, such as hemoglobin. Here,
there are additional parameter surfaces to be considered. In

CO Binding to Hemes
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4793
1
1
Figure 6. Z surfaces as a function of tilt and bend for the A0 substate of MbCO. (A) 13C-isotropic shift Z surface. (B) 17O-isotropic shift Z
1
1
13
13
17
17
surface. (C) 13C CSA Z surface. (D) 17O NQCC Z surface. (E)
C
,
iso
C
,
O
iso
,
O
4Z surface. The values given are the likelihoods
NQCC
CSA
n
n
x
Z.
shown in Figure 1A, and excellent accord was found for the
17O isotropic shift versus νCO correlation and a more modest
correlation for the 13C isotropic shift versus νCO correlation.14,36
We have now computed the 17O e2qQ/h versus νCO correlation,
using both uniform field and point changes to perturb the FeCO
fragment.69 We find derivatives ∂(e2qQ/h)/∂(νCO) of 11 kHz/
cm-1 for the uniform field model and 16 kHz/cm-1 for the point
charge model, to be compared with 11.28 kHz/cm-1 determined
experimentally.14,36,69 The averages of the uniform field and
point charge IR/NMR correlation lines are shown in Figure 8
and strongly support the idea that E-fields determine the major
changes seen between all the Ai substates, since the correlations
between 13Ci, 17Oi, 17O e2qQ/h, and νCO are well reproduced by
the calculations, although the absolute values of the properties
are not exactly the same as the protein or porphyrin results
because of the simplicity of the FeCO fragment.
Our results also suggest that E-field effects for the A0 substate
are close to zero, since there is no change in νCO in Fe(TPP)-
(CO)(NMeIm) in solution and in the crystalline solid state and
there is no distal fragment available to polarize the CO and
generate an E-field effect.
If these ideas are correct, it should therefore be possible to
extend the Bayesian probability approach to investigate the most
likely conformation of the A1 substate, simply by using the
known νCO to make a small correction to the observed 13C/17O
shifts and the 17O e2qQ/h values, using the theoretical correlation
lines shown in Figure 8. In addition, for the A1 substates of
heme proteins, there is an extensive database of 57Fe NMR
chemical shifts and 57Fe Mo¨ssbauer quadrupole splittings, ∆EQ,
which can also in principle be used to investigate τ,â effects.
We show therefore in Figure 9 computed 57Fe NMR isotropic
chemical shift and 57Fe Mo¨ssbauer quadrupole splitting surfaces
as a function of τ,â for a model Fe(P)(CO)(NMeIm) metallo-
cycle. The 57Fe NMR chemical shieldings, Figure 9A, com-
puted using the G94/DFT GIAO method described above are
given in Table 3 and have been converted into 57Fe NMR
chemical shifts, δi, in ppm from Fe(CO)5 using the following
equation
57Fe δ(shift, ppm from Fe(CO)5) ) -2922 - σ(calcd, ppm)
(3)
where σ(calcd, ppm) is the isotropic shielding given in Table
2. The value of -2922 ppm has been established by using the
databases of Bu¨hl (50) augmented with three model calculations
on metallocycles Fe(TPP) (Me2S)(NMeIm), Fe(TPP)(iPrNC)-
(NMeIm) and Fe(TPP)(CO)(NMeIm), using the experimental
shifts of cytochrome c,70 alkyl isocyanide myoglobins and
carbonmonoxymyoglobin, as will be described in detail else-
where.71
The Mo¨ssbauer quadrupole splittings are derived from the
electric-field gradient tensor, computed in Gaussian-94 using
the B3LYP hybrid exchange correlation functional. In Mo¨ss-
bauer spectroscopy, the observed spectra of low spin d6 iron
complexes generally consist of a quadrupole split doublet having
a peak separation, ∆EQ, which is related to the elements of the
electric field gradient tensor at the nucleus by
1/2
η2
3
∆EQ ) 1/2eQVzz 1 +
(4)
(
)
where e is the electron charge, Q the quadrupole moment of
the I* ) 3/2, 14.4 keV excited state, Vzz is the largest component
of the efg tensor, and by convention
(70) Baltzer, L. J. Am. Chem. Soc. 1987, 109, 3479-3481.
(71) Godbout, N.; Havlin, R.; Salzmann, R.; Debrunner, P. G. Oldfield,
E. J. Phys. Chem. A 1998, 102, 2342-2350.
(69) deDios, A. C.; Earle, E. Unpublished results.

4794 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
McMahon et al.
Table 6. Experimental Parameters for Fe(TPP)(CO)(NMeIm), A0 and A1 Myoglobins
13
17
system
13Ci (ppm)
C
(ppm)
17Oi (ppm)
O
(MHz)
57Fei (ppm)
8227e
∆EQ (mm/s)
IR (cm-1
)
CSA
NQCC
Fe(TPP)(CO)(NMeIm)
A0 myoglobin
A1 myoglobin
205.0a
205.5d
207.2d
453a
372a
372d
366a
1.0a
0.35b
1969c
1969d
1944
435a
447a
1.1d
0.8d
0.37f
a Determined in this work. b Reference 66. c Determined in Nujol. d Reference 14. e Reference 32. f Reference 22.
Vxx - Vyy
η )
(5)
(6)
Vzz
|Vzz| > |Vyy| > |Vxx|
In the presence of an applied field, each molecule has its efg
oriented differently in the field, and a powder spectrum is
observed. However, methods are available to evaluate such
powder averages as a function of Vii, from which ∆EQ and the
sign of ∆EQ can be deduced.72,73 Unfortunately, there has been
some debate as to the value of Q, the quadrupole moment of
the excited iron nucleus. Here, we use a recent and reportedly
very precise value, Q ) 0.16 × 10-28 cm2, to convert the
computed Vii values (in atomic units) to the Mo¨ssbauer
quadrupole splittings, ∆EQ (mm/s) using eq 4.
Elsewhere, we have investigated the 57Fe Mo¨ssbauer quad-
rupole splittings in 14 organometallic model systems66 and find
a <0.2 mm/s rms error between theory and experiment, which
gives an idea as to the accuracy with which ∆EQ can be
determined. Figure 9B shows the Mo¨ssbauer ∆EQ(τ,â) surface
computed for our model metalloporphyrin. In separate work,
we have found experimentally66 that the quadrupole splitting
of the model system Fe(5,10,15,20-tetraphenylporphyrin)(CO)-
(N-methylimidazole) is 0.35 mm/s, essentially the same as that
seen in MbCO, where values of 0.363-0.373 have been
reported.23,67 Large basis set efg calculations on the complete
TPP system yield a ∆EQ of 0.44 mm/s, to be compared with
our smaller model, having ∆EQ ) 0.52 mm/s.66 Since we know
that the linear Fe-C-O TPP model has a 0.35 mm/s ∆EQ, we
have used a -0.17 mm/s correction to the τ,â ∆EQ surface.
This is necessary, since otherwise we would conclude the TPP
model is slightly bent, which it is notsalthough we should note
that even an unscaled (0.52 mm/s τ ) â ) 0°) surface only
changes the final τ,â predictions by about 1°.
Figure 7. Experimental 13C and 17O solid-state MAS NMR spectra of
A1 CO horse myoglobin crystals at pH 7. (A) 13C CP-NQS NMR
spectrum at 90.5 MHz, -25 °C, 7 µs 90° pulse widths, 60 µs dephasing,
a 3 ms mix time, ωr ) 3.7 kHz, recycle ) 2 s, 88 944 scans, convolution
difference processing. (B) 67.8 MHz (11.7 T) 17O NMR spectrum of
C17O-myoglobin crystals. Experimental conditions were -24 °C,
1 048 000 scans, 50 ms recycle time, a 5.0 µs (solid 90°) pulse width,
and 25 Hz line broadening due to exponential multiplication. The
spinning sidebands are numbered.
1
For the iron-57 NMR chemical shift, the Z surface, Figure
9C, shows a most likely solution close to τ ) â ) 0°, although
given the 800 ppm rms error seen for three metalloprotein
models71 (as reflected in our 787 ppm value for W), clearly
considerable distortions are allowed, based solely on the 57Fe
NMR chemical shift. The same can be said for the iron-57
Mo¨ssbauer ∆EQ 1Z surface, Figure 9D. For example a τ ) â
) 20° solution is spectroscopically permissible, albeit not so
energetically. These results simply reinforce the idea that
multiple Z surfaces need to be combined in order to obtain a
τ,â solution which is consistent with all of the observables. We
therefore show in Figure 10 the 6Z surface obtained by
considering all of the NMR and Mo¨ssbauer data:
NMR chemical shift anisotropy, the isotropic 17O NMR chemi-
cal shift, the 17O nuclear quadrupole coupling constant, the 57Fe
NMR chemical shift, and the 57Fe Mo¨ssbauer quadrupole
splitting, as summarized in Table 7. Moreover, the result shown
in Figure 10 is not dominated by a single, narrow solution for
a particular spectroscopic parameter. For situations in which
there are extensive amounts of data, for example 13C NMR shifts
of amino acid residues in proteins, the W value chosen in eq 1
can be related to the rms or standard deviation of the data set.25,26
However, when only single-point experiments are available, W
values have to be deduced by estimating errors on a calculation
(and experiment) in the absence of a large body of data. In
such situations, a single badly chosen W value may dominate
the final Z surface. We have tested for this possibility by
selectively removing individual Z surfaces from the result shown
6Z(τ,â) ) 1Z(13C, δi)1Z(13C, CSA)1Z(17O, δi) ×
1Z(17O, e2qQ/h)1Z(57Fe,δi)1Z(57Fe, ∆EQ) (7)
5
The most likely solution occurs for τ ) 4°, â ) 7°, essentially
a linear and untilted Fe-C-O fragment. This result gives good
predictions for the isotropic 13C NMR chemical shift, the 13C
in Figure 10, to generate the six individual Z surfaces shown
in Figure 11. In all cases, the 5Z results are very similar to the
6Z results, with τ,â values very similar to the 4Z result obtained
with the A0 substate, Figure 6. That is, in no case does a single
Z surface dominate the final result. The actual W-values used,
together with a theoretical-versus-experimental comparison of
(72) Zimmerman, R. Nucl. Inst. Methods 1975, 128, 537-541.
(73) Mu¨nck, E.; Groves, J. L.; Tumolillo, T. A.; Debrunner, P. G.
Computer Phys. Commun. 1973, 5, 225-238.

CO Binding to Hemes
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4795
Figure 8. Experimental (solid circles, from ref 14) and theoretical (solid lines, from refs 36 and 69) NMR/IR correlation plots for heme proteins
and model compounds. (A) 13C-isotropic shift versus ν(C-O). (B) 17O-isotropic shift versus ν(C-O). (C) 17O NQCC versus ν(C-O). The solid
circles represent results obtained on proteins. The open triangles are the results obtained on the TPP model described in the text, and the open
square is C2-cap. The solid lines are the theoretical correlation lines, based on Fe-C-O model systems36,69 offset along the ordinate to intersect the
data at νCO ) 1969 cm-1, to facilitate comparison.
Figure 9. 57Fe NMR isotropic chemical shift and Mo¨ssbauer quadrupole splitting tilt-bend surfaces and associated Z surfaces for the A1 substate.
(A) 57Fe-isotropic shift parameter surface. (B) 57Fe quadrupole splitting parameter surface, offset by -0.17 mm/s to reproduce the Fe(TPP)(CO)-
1
1
(NMeIm) model system. (C) 57Fe-isotropic shift Z surface. (D) 57Fe quadrupole splitting Z surface. The values shown in A are the 57Fe shifts, in
103 ppm, and in B are the ∆EQ in mm/s. The values in C and D are the Z likelihoods.
1
all of the results obtained, are given in Table 7. Note that both
τ,â and E-field contributions are included. On average, the
mean 13C,17O ) isotropic shift error is 0.4 ppm, the 13C CSA
error is 7 ppm, the efg error is 0.13 MHz, the 57Fe ∆EQ error
is 0.05 mm/s, and the 57Fe-isotropic shift error is 179 ppm. For
the 13C and 17O shifts, as noted previously, the isotropic shifts
were scaled to fit the linear TPP and picket fence models, as
was the 57Fe Mo¨ssbauer surface. This is necessary in part to
account for basis and functional deficiencies and is similar to
the scaling we have used previously with amino acid shielding
in peptides26 and proteins.25 For the 13C CSA, the 17O NQCC,
and the 57Fe shifts, no scaling was used since the CSAs and
efgs were in good accord with the model compound experi-
ments, and for 57Fe, the solid-state 57Fe shift is unknown.
Indeed, the use of solely unscaled and uncorrected parameter
surfaces yields τ ) â ) 6°, in good agreement with the more
rigorous computations described above.
Figure 10. 6Z surface for A1 protein conformational substate using
13C δi, 17O δi, 57Fe δi, 13C ∆δ, 17O NQCC, and 57Fe ∆EQ surfaces. The
6
6
x
maximum is at τ ) 4°, â ) 7°. The maximum Z value is
Z )
0.92.

4796 J. Am. Chem. Soc., Vol. 120, No. 19, 1998
McMahon et al.
Figure 11. Six 5Z surfaces of A1 myoglobin illustrating the effect of removing each Z surface. A, 13C isotropic shift removed. B, 17O isotropic shift
removed. C, 57Fe isotropic shift removed. D, 13C CSA removed. E, 17O NQCC removed. F, 57Fe quadrupolar splitting removed. The countours
5
5
x
represent
Z and decrease in 0.1 intervals.
Table 7. Calculated and Experimental Spectroscopic Parameters
for Typical A0 and A1 CO-Heme Protein Substates
We believe these results therefore indicate that the A0 and
A1 substates of MbCO and, by inference other heme proteins,
contain linear Fe-C-O geometries, since all six spectroscopic
observables: the carbon-13, oxygen-17, and iron-57 NMR
isotropic chemical shifts, the 13C shift anisotropy, the 17O
quadrupole coupling constant, and the 57Fe Mo¨ssbauer quad-
rupole splitting, are all close to those found via DFT. This
improves upon the more limited results using just DFT energies
alone or solely IR data, which appear to leave some uncertainties
as to the actual τ,â values present in proteins, with in some
cases large distortions reported as possible (see, e.g., ref 74).
The changes in the ligand NMR/IR shifts from system to system
seen previously can be readily accounted for in terms of E-field
effects, again using DFT methods. These results also suggest
that the A3 substate is very close to linear and untilted, as are
the unusually shifted peroxidase “post-A3” substates, since (i)
all points fall on the universal δi 17O/δi 13C/17O NQCC/νCO
correlations, (ii) the slopes of these correlations can now be
computed via use of DFT methods, and (iii) these correlations
are not due to τ,â.
While an argument might be made that the highly distorted
Fe-C-O structures are only found in P21 sperm whale MbCO,
the Mo¨ssbauer surface results give no support to this idea, and
indeed DFT calculations at the X-ray geometries overestimate
∆EQ by 500%,66 while assumption of a linear geometry gives
results within the 0.1-0.2 mm/s rms error expected. Indeed,
the fact that both solution and solid-state data can all be
combined together using the Bayesian probability approach to
give unique solutions, which permit an explanation of seven
different observables from three different spectroscopies, NMR,
IR, and Mo¨ssbauer, which also cover a dozen or more proteins
and model systems (C2Cap, A0, A1, A3, peroxidases), gives some
confidence in the use of such methods for investigating related
structural questions in other systems as well.
A0
A1
parameter
calcd exptl
W
calcd
exptla
W
13
13
17
17
C
C
O
O
(ppm)
(ppm)
(ppm)
205
446
372
205.3
435
372.4
1.1
3.2
28.0
4.5
206.8
445
366.7
0.64
0.30
8406
207.2
447
366
0.8
0.37
8227
3.2
28.0
4.5
0.6
0.4
iso
CSA
iso
(MHz)
0.9
0.6
NQCC
57Fe∆EQ (mm/s)
57Feiso (ppm)
787
a The electrical contributions to 13C δi, 17O δi and the 17O NQCC
were obtained from the theoretical correlation lines shown in Figure
8.
In summary then, the use of density functional theory now
permits the relatively accurate prediction of 13C NMR chemical
shifts and shift tensor elements, 17O NMR shifts, 17O nuclear
quadrupole coupling constants, 57Fe NMR chemical shifts and
57Fe Mo¨ssbauer quadrupole splittings in metalloporphyrin model
systems and in metalloproteins themselves. Experimental results
on linear and untilted Fe-C-O metalloporphyrins are es-
sentially the same as those found in the A0 substate of heme
proteins. This implies that the proteins could also contain linear
and untilted Fe-C-O units, and the results of additional DFT
calculations rule out alternative, distorted conformations for the
A0 substate. With the A1 substates of heme proteins, there are
small changes in the ligand NMR and electric field gradient
parameters, but these can now be rather well described,
especially for the δi 17O and 17O NQCC/νCO correlations, in
terms of purely electrostatic field effects, calculable via quantum
chemistry. For the A1 substate, we also have available the 57Fe
NMR isotropic shifts and the 57Fe Mo¨ssbauer quadrupole
splittings, ∆EQ. These are not expected to change appreciably
with E-field effects, since these in general come from the distal
histidine and as such are expected to affect primarily the 17O
shift and 17O NQCC. The 57Fe shift and 57Fe efg can therefore
be used as additional parameters with which to test ideas about
metal-ligand geometry. Using the simple porphyrin model,
we find errors of about 200 ppm for the 57Fe shift (8227 ppm
expt, 8406 ppm calcd), a small number when compared with
the ∼1000-3000 ppm errors found for 20°, 40° bent geometries.
The same can be said for the 57Fe Mo¨ssbauer results, where-
highly bent geometries result in very large errors in computed
57Fe efgs, while linear and untilted geometries give computed
Mo¨ssbauer quadrupole splittings, ∆EQ, which are very close to
those found experimentally.
Conclusions
The results we have shown above are of interest for three
reasons. First, they represent the first comprehensive investiga-
tion of 13C NMR chemical shifts, 13C NMR chemical shift
(74) Spiro, T. G.; Kozlowski, P. M. J. Biol. Inorg. Chem. 1997, 2, 516-
520. Slebodnick, C.; Ibers, J. A. J. Biol. Inorg. Chem. 1997, 2, 521-525.
Vangberg, T.; Bocian, D. F.; Ghosh, A. J. Biol. Inorg. Chem. 1997, 2, 526-
530. Sage, T. J. J. Biol. Inorg. Chem. 1997, 2, 537-543.
(75) Ries, W.; Bernal, I.; Quast, M.; Albright, T. A. Inorg. Chim. Acta
A 1984, 83, 5-15.

CO Binding to Hemes
J. Am. Chem. Soc., Vol. 120, No. 19, 1998 4797
anisotropies, 17O NMR chemical shifts, 17O nuclear quadrupole
couplings, 57Fe NMR chemical shifts, and 57Fe Mo¨ssbauer
quadrupole splittings in metalloproteins and model systems using
modern density functional methods and experiment. Second,
our results indicate that all of the spectroscopic observables can
be well accounted for in terms of linear, untilted Fe-C-O
bonding, with weak electrostatic fields controlling the differ-
ences seen between A0 and A1 substates. Third, we have
extended the use of the Z surface method, based on Bayesian
probability, to explore the topic of ligand tilt and bend: the
results give no support for distorted geometries, either in solution
or in the solid-state. Combining the results of NMR, IR, and
Mo¨ssbauer spectroscopic measurements in this way should be
a general approach for predicting metal-ligand geometries in
other systems as well, for example with O2 and RNC metal-
ligand bonding in heme proteins, where the more conventional
IR/Raman-based methods alone are less sensitive.
Institute grant HL-19481). This work was also supported in part
by use of the SGI/Cray Origin 2000 and Power Challenge
clusters at the National Center for Supercomputing Applications
(funded in part by the US National Science Foundation grant
CHE-970020N). We also gratefully acknowledge Professor
Dennis Salahub, Drs. Vladimir Malkin and Olga Malkina for
providing a copy of their deMon and SOS-DFPT deMon
program, and Professor U. Nienhaus for initial FT-IR measure-
ments. M.T.M. is a National Institutes of Health Molecular
Biophysics Training Grant Trainee, A.C.D. is an American Heart
Association, Inc., Illinois Affiliate, Postdoctoral Fellow (1994-
1995). R.S. is a Swiss National Science Foundation Postdoctoral
Fellow (1996-1997) and an American Heart Association, Inc.,
Illinois Affiliate, Postdoctoral Fellow (1997-1998). D.D.L. is
a Colgate-Palmolive Fellow. R.H.H. is a Barry Goldwater
Fellow.
Acknowledgment. This work was supported by the United
JA973272J
States Public Health Service (National Heart, Lung and Blood