Differential sensing of protein influences by NO and CO vibrations in heme adducts

Article abstract of DOI:10.1021/ja064859d

Heme proteins bind the gaseous ligands XO (X = C, N, O) via backbonding from Fe dπ electrons. Backbonding is modulated by distal interactions of the bound ligand with the surrounding protein and by variations in the strength of the trans proximal ligand. Vibrational modes associated with FeX and XO bond stretching coordinates report on these interactions, but the interpretive framework developed for CO adducts, involving anticorrelations of νFeC and νCO, has seemed not to apply to NO adducts. We have now obtained an excellent anticorrelation of νFeN and νNO, via resonance Raman spectroscopy on (N-methylimidazole)Fe(II)TPP-Y(NO), where TPP-Y is tetraphenylporphine with electron-donating or -withdrawing substituents, Y, that modulate the backbonding; the problem of laser-induced dissociation of the axial base was circumvented by using frozen solutions. New data are also reported for CO adducts. The anticorrelations are supported by DFT calculations of structures and spectra. When protein data are examined, the NO adducts show large deviations from the modeled anticorrelation when there are distal H-bonds or positive charges. These deviations are proposed to result from closing of the FeNO angle due to a shift in the valence isomer equilibrium toward the Fe(III)(NO^-) form, an effect that is absent in CO adducts. The differing vibrational patterns of CO and NO adducts provide complementary information with respect to protein interactions, which may help to elucidate the mechanisms of ligand discrimination and signaling in heme sensor proteins.

Full text of DOI:10.1021/ja064859d

Published on Web 12/09/2006
Differential Sensing of Protein Influences by NO and CO
Vibrations in Heme Adducts
Mohammed Ibrahim, Changliang Xu, and Thomas G. Spiro*
Contribution from the Department of Chemistry, Princeton UniVersity,
Princeton, New Jersey 08544
Received July 16, 2006; E-mail: spiro@princeton.edu
Abstract: Heme proteins bind the gaseous ligands XO (X ) C, N, O) via backbonding from Fe dπ electrons.
Backbonding is modulated by distal interactions of the bound ligand with the surrounding protein and by
variations in the strength of the trans proximal ligand. Vibrational modes associated with FeX and XO
bond stretching coordinates report on these interactions, but the interpretive framework developed for CO
adducts, involving anticorrelations of νFeC and νCO, has seemed not to apply to NO adducts. We have
now obtained an excellent anticorrelation of νFeN and νNO, via resonance Raman spectroscopy on (N-
methylimidazole)Fe(II)TPP-Y(NO), where TPP-Y is tetraphenylporphine with electron-donating or
-withdrawing substituents, Y, that modulate the backbonding; the problem of laser-induced dissociation of
the axial base was circumvented by using frozen solutions. New data are also reported for CO adducts.
The anticorrelations are supported by DFT calculations of structures and spectra. When protein data are
examined, the NO adducts show large deviations from the modeled anticorrelation when there are distal
H-bonds or positive charges. These deviations are proposed to result from closing of the FeNO angle due
to a shift in the valence isomer equilibrium toward the Fe(III)(NO-) form, an effect that is absent in CO
adducts. The differing vibrational patterns of CO and NO adducts provide complementary information with
respect to protein interactions, which may help to elucidate the mechanisms of ligand discrimination and
signaling in heme sensor proteins.
Introduction
linear correlations between the X-O stretching and Fe-X
stretching frequencies have been observed for Fe(II) porphyrin
The gaseous molecules CO, NO, and O2 (XO) bind avidly
to heme, thanks to effective backdonation of Fe(II) dπ electrons
to their π* orbitals. Nature has harnessed this property to
regulate essential biochemical processes in response to the
presence of these molecules. Heme sensor proteins are the signal
proteins for these regulatory mechanisms.1 They contain an
enzymatic or DNA-binding domain, linked to a heme domain.
Activity is turned on or off when the appropriate ligand binds
to the heme. There is great interest in understanding how the
heme domain discriminates among the three XO molecules, and
how the binding event is transduced into activation.
adducts with all three X-O ligands.2,3 In the case of CO,
negative backbonding correlations are seen as well for six-
coordinate adducts with axial base ligands.3 However, the
correlations are displaced to higher or lower νCO when the axial
ligand bond is weakened or strengthened, because the effect
on the Fe-C bond due to changes in backbonding is compen-
sated by σ bonding competition, leaving νFeC essentially
unaltered.4
These correlations are obeyed in CO adducts of heme
proteins, as well as protein-free complexes, and have been used
extensively to gauge the nature of distal polar interactions, which
affect backbonding, or the weakening of the axial ligand bond
via heme displacement.4-6 However, NO adducts of heme
proteins are less well behaved. Plots of νNO against νFeN are
scattered, and the variation in νFeN is small.7-9 This behavior
led Park and Boxer to suggest that backbonding is unimportant
Both phenomena, discrimination and signaling, involve
changes in the interactions between the heme group and the
surrounding protein when the XO ligand binds. Resonance
Raman spectroscopy can provide useful information about these
interactions, because it can separately monitor vibrational modes
of the protein, the porphyrin macrocycle, and the heme-bound
ligand.
(2) Vogel, K. M.; Kozlowski, P. M.; Zgierski, M. Z.; Spiro, T. G. J. Am. Chem.
Soc. 1999, 121, 9915-9921.
Backdonation produces anticorrelation between the X-O and
Fe-X bond strengths; as backdonation increases the Fe-X bond
order increases, while the X-O bond order decreases. Negative
(3) Vogel, K. M.; Kozlowski, P. M.; Zgierski, M. Z.; Spiro, T. G. Inorg. Chim.
Acta 2000, 297, 11-17.
(4) Spiro, T. G.; Wasbotten, I. H. J. Inorg. Biochem. 2005, 99, 34-44.
(5) Coyle, C. M.; Puranik, M.; Youn, H.; Nielsen, S. B.; Williams, R. D.;
Kerby, R. L.; Roberts, G. P.; Spiro, T. G. J. Biol. Chem. 2003, 278, 35384-
35393.
(1) (a) Rodgers, K. R. Curr. Opin. Chem. Biol. 1999, 3, 158-67. (b) Gilles-
Gonzalez, M. A.; Gonzalez, G. J. Inorg. Biochem. 2005, 99, 1-22. (c)
Boon, E. M.; Marletta, M. A. Curr. Opin. Chem. Biol. 2005, 9, 441-446.
(d) Roberts, G. P.; Kerby, R. L.; Youn, H.; Conrad, M. J. Inorg. Biochem.
2005, 99, 280-292.
(6) Ibrahim, M.; Kerby, R. L.; Puranik, M.; Wasbotten, I. H.; Youn, H.; Roberts,
G. P.; Spiro, T. G. J. Biol. Chem. 2006, 281, 29165-29173.
(7) Thomas, M. R.; Brown, D.; Franzen, S.; Boxer, S. G. Biochemistry 2001,
40, 15047-15056.
(8) Park, E. S.; Boxer, S. G. J. Phys. Chem. B 2002, 106, 5800-5806.
9
16834
J. AM. CHEM. SOC. 2006, 128, 16834-16845
10.1021/ja064859d CCC: $33.50 © 2006 American Chemical Society

NO and CO Vibrations in Heme Adducts
A R T I C L E S
in heme-NO adducts and that protein-induced electrostatic
shifts in νNO (Stark effect) are associated with anharmonicity
in the NO stretch.8 As noted above, however, νNO and νFeN
are anticorrelated for five-coordinate (5-c) Fe(II) porphyrin NO
adducts, with a slope essentially the same as that observed for
the corresponding CO adducts.2 At least in these adducts,
backbonding is as important for NO as it is for CO. On the
other hand, vibrational data have not been available for protein-
free six-coordinate (6-c) NO adducts, leaving open the pos-
sibility that backbonding is somehow vitiated by the axial ligand.
Because of the strong trans effect in Fe(II)NO adducts, axial
ligands bind weakly. We found it impossible to obtain 6-c RR
spectra even at axial base concentrations sufficient to observe
6-c UV/vis absorption spectra, and we concluded that the axial
ligand is dissociated by photo and/or thermal effects of the
Raman laser.9
Another ambiguity concerns the nature of the vibrations
themselves. From nuclear resonance vibrational spectroscopy
(NRVS), Sage and co-workers10 found that the mode commonly
assigned to νFeN in MbNO at 556 cm-1 actually has a smaller
Fe displacement than does another mode, at 451 cm-1, which
they suggested should be reassigned to νFeN. Could the
correlation problem lie in tracking the wrong mode? Because
the FeNO equilibrium geometry is bent (∼140°), the Fe-N
stretching and Fe-N-O bending coordinates are necessarily
combined into in- and out-of-phase combinations in the relevant
normal modes. (The FeCO unit is essentially linear, so the
mixing problem does not arise.) How much each contributes is
a quantitative matter.
In this study, we re-examine these issues with fresh computa-
tions, and with new experimental data. On the computational
side, we find, as have others,11 that the adequacy of DFT
calculations is sensitive to the nature of the functional. In the
case of 6-c NO adducts, the σ bonding competition between
NO and the trans axial ligand is difficult to capture accurately
because of the strong trans labilizing effect of NO.2 In this
regard, we find that hybrid functionals do a poorer job than
non-hybrid functionals, probably by introducing spin contamina-
tion. The best agreement with experimental bond-distances is
given by the non-hybrid BLYP functional, and the normal modes
extracted from this calculation have Fe-N stretching as the
dominant coordinate for the higher of the stretch/bend modes,
for both 5-c and 6-c adducts, justifying the conventional
experimental assignment of νFeN.
appending electron-donor and -acceptor substituents to porphine
in silico, although the results are less satisfactory for NO than
for CO, reflecting the greater electronic complexity of the NO
adducts.
When protein data are re-examined in the light of these
results, it becomes evident that the scatterplots seen for NO
adducts reflect deviations from the expected anticorrelation that
depend on the presence of distal H-bonds or positive charges.
A bonding model is developed in which these electrostatic
effects diminish the FeNO angle by stabilizing the Fe(III)(NO-)
valence isomer. This model may prove useful in mapping the
heme pocket interactions of sensor proteins, when both CO and
NO data are considered.
Methods
Materials. Iron(III) tetraphenylporphyrin chloride, Fe(III)(TPP)Cl,
and its phenyl-substituted analogues (Fe(III)(TPP-Y)Cl, where Y )
p-hydroxy, p-methoxy, p-methyl, 2,6-dicholoro, 2,6-difluoro, p-nitro,
and pentafluoro, respectively) were purchased from Midcentury Chemi-
cals (Posen, IL). Some of these samples that possess fluorescing
impurities were purified by treating the sample solutions with activated
charcoal (EM Science, Cherry Hill, NJ). N-Methylimidazole (N-MeIm)
(99+ %) and methanol (99.8%, anhydrous) were purchased from
Aldrich. Methylene chloride (Optima) was purchased from Fisher
Scientific. N,N-Dimethylformamide (DMF, spectrophotometric grade)
was obtained from J. T. Baker (Phillipsburg, NJ).
Preparation of the 6-c Fe(II)-NO Complexes. 6-c NO adducts,
(N-MeIm)Fe(II)TPP-Y(NO), were prepared according to the literature
procedure12 with some modifications. Specifically, to a mixture of 110
µL of N-MeIm (1.4 mmol) and 15 µL methanol (0.37 mmol) in an
EPR tube was added iron(III) porphyrin stock solution in CH2Cl2 to
make a final concentration of ∼400 µM and a volume of 150 µL. The
EPR tube was sealed with a septum, and the solution was deoxygenated
by flushing with pure nitrogen gas for ∼15 min. Next, 14NO gas (C.P,
Matheson Gas Products Inc.), which was prepurified by bubbling
through a 2.5 M NaOH solution and water, was introduced to the sample
for 5-10 min. The formation of 6-c FeII-14NO was checked and
confirmed by absorption spectroscopy (via observing subtle changes
in the Q bands).
The 6-c FeII-15NO complexes were prepared in a similar fashion,
except that nitric oxide 15NO gas was generated by reacting sodium
nitrite, Na15NO2 (15N 99%, Icon Services, Summit, NJ), with L-ascorbic
acid (Aldrich) in an anaerobic aqueous solution. The gas generated
was immediately transferred, via a gastight syringe, to the degassed
porphyrin solution in the EPR tube, and formation of 6-c NO adduct
was checked with absorption spectroscopy.
Preparation of 6-c Fe(II)-CO Complexes. To a 5 µL of N-MeIm
(63 µmol) in an EPR tube was added Fe(III) porphyrin solution (in
DMF or CH2Cl2) to make a final concentration of ∼400 µM and a
volume of 150 µL. The tube was then sealed and purged with N2 for
∼15 min. The sample was reduced by addition of 17 mM (final
concentration) of an aqueous anaerobic solution of sodium dithionite
(Na2S2O4, Aldrich) via a gastight syringe. 12CO gas (BOC Gases) was
flushed into the solution for ∼10 min to form Fe(II)-CO adducts. For
the 13CO analogue, the EPR tube was connected to the vacuum system,
purged with N2 gas for 15 min, and then evacuated; this step was
repeated three times, and then 13CO gas (13C 99%, Icon Services,
Summit, NJ) was introduced to the sample. The formation of the 6-c
FeII-CO complexes was confirmed by absorption spectroscopy.
Raman Spectroscopy. RR spectra were collected at 77 K by using
the 413.1 or 406.7 nm lines of the Kr ion-laser (2080-RS, Spectra
Physics) via backscattering geometry. Photodissociation of bound
On the experimental side, we have overcome the problem of
axial base lability by freezing solutions of 6-c adducts. The
frozen solution enforces efficient recombination of the photo-
detached axial ligand, permitting 6-c RR spectra to be recorded.
Data on 6-c adducts for a series of porphyrins having electron
donor or acceptor peripheral substituents produce a well-behaved
anti-correlation between NO and FeN, just as for CO adducts,
albeit with different slope and different displacement from the
5-c line. Decent backbonding correlations can be computed by
(9) Coyle, C. M.; Vogel, K. M.; Rush, T. S., III; Kozlowski, P. M.; Williams,
R.; Spiro, T. G.; Dou, Y.; Ikeda-Saito, M.; Olson, J. S.; Zgierski, M. Z.
Biochemistry 2003, 42, 4896-4903.
(10) Zeng, W.; Silvernail, N. J.; Wharton, D. C.; Georgiev, G. Y.; Leu, B. M.;
Scheidt, W. R.; Zhao, J.; Sturhahn, W.; Alp, E. E.; Sage, J. T. J. Am. Chem.
Soc. 2005, 127, 11200-11201.
(11) (a) Zhang, Y.; Gossman, W.; Oldfield, E. J. Am. Chem. Soc. 2003, 125,
16387-16396. (b) Praneeth, V. K. K.; Neese, F.; Lehnert, N. Inorg. Chem.
2005, 44, 2570-2572. (c) Tangen, E.; Svadberg, A.; Ghosh, A. Inorg.
Chem. 2005, 44, 7802-7805.
(12) Stong, J. D.; Burke, J. M.; Daly, P.; Wright, P.; Spiro, T. G. J. Am. Chem.
Soc. 1980, 102, 5815-5819.
9
J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006 16835

A R T I C L E S
Ibrahim et al.
Figure 1. RR spectra (413.1 nm excitation) for frozen solutions (77 K) of (N-MeIm)Fe(II)T(2,6-F2P)P(NO) containing 14NO or 15NO, and the difference
spectra.
N-MeIm, NO, and CO was minimized by using low laser power
(∼0.7-1 mW) at the sample. The spectra were collected with a triple
monochromator (Spex 1877 Triplimate, Spex Industries, Inc.) and a
liquid nitrogen cooled CCD detector (model 7375-0001, Roper
Scientific) operating at -110 °C. Spectra were calibrated with dimethyl
formamide, ethyl acetate, and dimethyl sulfoxide-d6 and analyzed by
Grams/AI software (version 7.0, Thermo-Galactic).
To overcome these problems, we resorted to freezing the
sample, hoping that the frozen solvent matrix would induce rapid
and efficient recombination of photodissociated ligand. This
strategy succeeded, as illustrated in Figure 1. The RR spectrum
of a frozen solution of (N-MeIm)Fe(II)T(2,6-F2P)P(NO) in
methylene chloride, to which ∼9 M (final concentration)
N-methylimidazole (N-MeIm) had been added, clearly showed
νFeN and νNO bands at frequencies in the range of values seen
for Fe(II)NO adducts of heme proteins, and very different from
the values seen for 5-c adducts. The 1632 cm-1 νNO band is a
shoulder on the side of a stronger porphyrin band, but the latter
is subtracted out in the 14NO-15NO difference spectrum, which
reveals the expected 32 cm-1 isotope shift. The 574 cm-1 νFeN
band is more prominent, but is in a crowded region of the
spectrum. The 14NO-15NO difference spectrum reveals a double
band in the 15NO spectrum, indicating that the νFeN mode has
shifted into sufficiently close coincidence with a porphyrin mode
to produce a Fermi resonance, in which intensity is shared
between the two modes. Fermi resonances were previously
encountered for the νNO RR band of 6-c Fe(II)NO adducts of
Mb variants,9 whose higher νNO frequency falls in another
crowded region of porphyrin modes. As in that case, we
determined an effective mode frequency by curve fitting the
pair of RR difference bands, and averaging their frequencies.
The resulting frequency for the 15NO isotopomer gave the
expected ∼20 cm-1 isotope shift.
Computations. Different DFT functionals were used in this study
to better reproduce the experimental observations: B3LYP for heme-
CO models and BLYP for heme-NO models (other DFT functionals
were applied to heme-NO models as well; results in Table S1). The
standard 6-31G* basis set was used for all of the atoms except Fe, on
which Ahlrichs’ valence triple-ú (VTZ)13,14 basis set was employed.
Calculations of geometry optimization and frequencies were performed
with the Gaussian 03 program15 applied with tight convergence criteria
and an ultrafine integration grid.
All of the frequency values were taken directly from the Gaussian
program without scaling. Normal mode and isotopic shift analyses were
performed by the TX9016 program based on Cartesian force constants
calculated at the optimized geometry.
Results and Discussion
Raman Spectra of 6-c NO and CO Adducts. A principal
objective of this work was to investigate the extent of back-
bonding in 6-coordinate (6-c) Fe(II)NO porphyrin adducts from
the νFeN and νNO frequencies. RR spectra have been reported
for such adducts only in heme proteins. Without the protein
constraints, the trans axial ligand is too labile to permit
acquisition of RR spectra in fluid solution. In our experience,
even when sufficient ligand is added to ensure complete
formation of the 6-c adduct, as judged by the UV/vis absorption
spectrum, the RR spectrum is that of the 5-c adduct. Indeed,
published reports of RR spectra of 6-c adducts have frequencies
that are characteristic of 5-c adducts.17,18 We infer that the
Raman laser induces dissociation of the weak Fe-ligand bond,
either photolytically or via local heating.
As in the previous 5-c adduct study,2 we varied the extent of
backdonation in a series of tetraphenylporphine adducts, having
phenyl substituents, Y, with a range of electron-donating or
-accepting properties. The isotope difference spectra are shown
in Figure 2, and the frequencies are collected in Table 1. The
νNO frequency increases systematically from the most (p-
OCH3P) to the least (F5P) electron-donating phenyl substituent,
while νFeN decreases in the same series. This pattern is the
expected signature of backbonding.
Because protein-free 6-c Fe(II)CO adducts have not previ-
ously been studied systematically, we measured their RR spectra
(Figures 3 and 4) as the phenyl-Y substituents were varied.
Fewer representatives were available, because the electron-
withdrawing substituents tended to increase the photolability
of CO, making acquisition of RR spectra difficult. Nevertheless,
(13) Schafer, A.; Horn, H.; Ahlrichs, R. J. Chem. Phys. 1992, 97, 2571-2577.
(14) Bauernschmitt, R.; Ahlrichs, R. Chem. Phys. Lett. 1996, 256, 454-464.
(15) Frisch, M. J.; et al. Gaussian 03; Gaussian, Inc.: Wallingford, CT, 2004.
(16) Pulay, P. TX90; Fayetteville, AR, 1990. Pulay, P. Theor. Chim. Acta 1979,
50, 299-312.
(17) Lipscomb, L. A.; Lee, B.-S.; Yu, N.-T. Inorg. Chem. 1993, 32, 281-286.
(18) Praneeth, V. K. K.; Nather, C.; Peters, G.; Lehnert, N. Inorg. Chem. 2006,
45, 2795-2811.
9
16836 J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006

NO and CO Vibrations in Heme Adducts
A R T I C L E S
Figure 2. RR difference spectra (14NO-15NO) of (N-MeIm)Fe(II)(TPP-Y)(NO) with the indicated phenyl substituent, Y. The dotted lines are the results
of curve fitting (50% Gaussian/50% Lorentzian, with 10 cm-1 bandwidths) to resolve the 15NO Fermi resonance at ∼550 cm-1
.
Table 1. Experimental Vibrational Frequencies and Isotopic Shifts for (N-MeIm)FeII(TPP-Y)(L)(XO) Complexes with Indicated Phenyl
Substituent, Y
substituent (Y)
1
1
1
1
(solvent)
ν
(Fe−
N) (cm-
)
∆
15N
ν(Fe
−
C) (cm-
)
∆
13C
ν(N
−
O) (cm-
)
∆
15N
ν
(C−
O) (cm-
)
∆
13C
p-OH (DMF)
501
498
4
3
1953
1960
45
45
p-OCH3 (CH2Cl2)
p-CH3 (CH2Cl2)
H(CH2Cl2)
2,6-Cl2 (CH2Cl2)
2,6-F2 (CH2Cl2)
F5 (CH2Cl2)
587
586
582
576
574
577
22
1621
1622
1623
1630
1632
1632
33
18
15
23
21
23
31
32
30
32
29
495
3
1962
46
p-NO2 (DMF)
492
3
1966
45
Figure 3. RR spectra (413.1 nm excitation) for frozen solutions (77 K) of (N-MeIm)Fe(II)T(p-OH-P)P(CO) containing 12CO or 13CO, and the difference
spectra.
four pairs of νCO/νFeC frequencies (Table 1) were obtained,
sufficient to establish the backbonding correlation (see next
section).
is the range of variation in both sets of frequencies that are
induced by the differing phenyl-Y substituents.
Negative correlations are obtained for all four sets of data,
reflecting Fe-XO backbonding. As noted earlier,2 the slopes
of the 5-c adduct plots are nearly the same for CO (-0.46) and
NO (-0.40). The slope increases for the 6-c adducts, reflecting
an increased sensitivity of νFeX for a given backbonding
increment in νXO.3 The increase is even larger for NO (-1.0)
than for CO (-0.68). Thus, the sensitivity for νFeN is especially
TPP-X Backbonding Correlations. Figure 5 shows νFeX/
νXO plots for 5- and 6-c Fe(II)porphyrin adducts with CO and
NO. There is a remarkable parallelism, despite the different
electronic structures of NO and CO. The νCO and νNO values
are, of course, very different, reflecting the extra antibonding
electron on NO, but the νFeC and νFeN values are similar, as
9
J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006 16837

A R T I C L E S
Ibrahim et al.
Table 2. Backbonding Parametersa for Fe(II) Porphyrin NO and
CO Adducts
1
1
ν°(Fe−
C) (cm-
)
slope (s)
ν°(Fe−
N) (cm-
)
slope (s)
Experimental
-0.46
five-coordinate
six-coordinate
435
371
445
329
-0.40
-1.0
-0.68
DFT-Calculated
-0.58
five-coordinate
six-coordinate
534
443
-0.82
523
-0.37
a νFeX ) ν°FeX - s[νXO - ν°XO]; ν°XO (gas phase) ) 2145 cm-1 for CO
and 1876 cm-1 for NO.
cm-1) and NO (1876 cm-1), and ν°FeX is the limiting (single-
bond) frequency in the absence of backbonding. The parameters
are collected in Table 2, where it can be seen that the ν°FeX
limit values are quite similar for 5-c adducts and are greatly
depressed in both cases by the addition of axial ligand. The 6-c
νFeN elevation seen for the actual NO complexes disappears
for the limit value, because of the large 6-c slope. The limit
values are, of course, entirely artificial, because backbonding
is integral to the FeXO electronic structures.
Figure 4. RR spectra (low-frequency region, left) and difference spectra
(12CO-13CO) (high-frequency region, right) of (N-MeIm)Fe(II)(TPP-Y)-
(CO) with the indicated phenyl substituent, Y. (The Fe-C band itself is
shown instead of the isotope difference band, whose positive and negative
lobes do not accurately represent mode frequencies in this case because of
the small isotope shift (ref 19).)
DFT Computation of Structure. Our DFT-computed struc-
tures agree well with experimental data (Table 3 and Figure 6).
For the NO adducts, the bonds to the porphyrin N atoms are
correctly predicted to be shorter in the direction of Fe-N-O
bending. (This point is discussed in the crystal structure
papers,20,21 although the individual distances have been averaged
in the reported distances, Table 3.)
The structures reveal the different responses of the CO and
NO adducts to the addition of an imidazole ligand. In each case,
the Fe-X bond is lengthened, but the displacement is twice as
large for CO (0.06 Å) as for NO (0.03 Å); theory and experiment
agree on these displacements. In contrast, the Fe-ImH bond is
longer for NO (2.24 Å) than for CO (2.08 Å): this is the well-
known NO trans effect. Because the bond is weak, experimental
bond distances for the bond trans to NO are somewhat variable,
and Scheidt has not given a “canonical” distance.21 The
computed Fe-ImH distance is close to that reported for the
Fe-pyridine distance in (Py)Fe(TpivPP)(NO) (2.26 Å)21 and
somewhat longer than the Fe-N-methylimidazole distance in
(N-MeIm)Fe(TPP)(NO) (2.17 Å).21 For the CO adduct, the
computed Fe-ImH distance is close to the experimental value
for (N-MeIm)Fe(TPP)(CO) (2.05 Å)22 and is somewhat longer
than the 2.00 Å distance reported for the bis-imidazole adduct.21
The picture that emerges is that the trans ligand competes
for bonding to Fe with either CO or NO. In both cases, the
Fe-ImH and Fe-X bonds are both longer than if the opposite
ligand is absent. However, the competition is fairly even for
CO (∼0.06 Å bond extension for both Fe-CO and Fe-ImH),
but decidedly uneven for NO (0.03 Å extension for Fe-NO
and 0.16 Å for Fe-ImH).
Figure 5. FeXO backbonding correlations for 5-c and 6-c (N-MeIm axial
ligand) NO and CO adducts of Fe(II)TPP-Y with the indicated phenyl
substituents, Y, in organic solvents (CH2Cl2, DMF, Bz). For the 5-c adducts,
NO data are from ref 2 and CO data are from ref 3. The 6-c adduct data
are from Table 1.
high, despite the impression gained from a set of myoglobin
variants that the range of νFeN is small (see below).
The main difference between the CO and NO plots is that
the 6-c line lies below the 5-c line for CO but above the 5-c
line for NO. Adding an axial ligand depresses the νFeC
frequency but elevates the νFeN frequency. There are two
contributing factors to this difference, as discussed in the
following section: (i) The axial ligand lengthens the Fe-X bond
for both Fe-C and Fe-N, but less so for Fe-N, due to the
strong NO trans effect. (ii) Because the Fe-N-O angle is bent,
the Fe-N stretching coordinate mixes with the Fe-N-O
bending coordinate, and this mixing increases when an axial
ligand is bound.
For the CO adducts, the widely used hybrid functional B3LYP
serves very well, but we found that it gave poor results for the
(19) Song, S.; Boffi, A.; Chiancone, E.; Rousseau, D. L. Biochemistry 1993,
32, 6330 - 6336.
(20) Scheidt, W. R.; Haller, K. J.; Fons, M.; Mashiko, T.; Reed, C. A.
Biochemistry 1981, 20, 3653-3657.
The backbonding correlations can be cast in a standard form,3
for ease of comparison:
(21) Wyllie, G. R. A.; Schulz, C. E.; Scheidt, W. R. Inorg. Chem. 2003, 42,
5722-5734.
(22) Silvernail, N. J.; Roth, A.; Schulz, C. E.; Noll, B. C.; Scheidt, W. R. J.
Am. Chem. Soc. 2005, 127, 14422-14433.
νFeX ) ν°FeX - s[νXO - ν°XO]
(23) Scheidt, W. R.; Frisse, M. E. J. Am. Chem. Soc. 1975, 97, 17-21.
(24) Nasri, H.; Haller, K. J.; Wang, Y.; Hanh, H. B.; Scheidt, W. R. Inorg.
Chem. 1992, 31, 3459 - 3467.
where ν°XO is the frequency for unbound (gas phase) CO (2145
9
16838 J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006

NO and CO Vibrations in Heme Adducts
A R T I C L E S
Table 3. Experimental and Computed Structural Parameters for 5-c and 6-c FeXO Adducts
compound
Fe
−XO (Å)
X
−O (Å)
Fe
−X
−O (deg)
Fe
−
L (Å)a
Fe−
Np (Å)b
ref
Five-Coordinate CO
Fe(Deut)(CO)(THF)c
calc:d FeP(CO)
1.706(5)
1.735
1.144(5)
1.152
177.4(9)
180.0
2.127(4)
1.980
2.008
20
t.w.
Six-Coordinate CO
(N-MeIm)Fe(TPP)(CO)e
calc:d (ImH)FeP(CO)
1.7600(17)
1.796
1.139(2)
1.150
177.03(15)
180.0
2.0503(14)
2.082
2.005(6)
2.024/2.026
22
t.w.
Five-Coordinate NO
Fe(TPP)(NO)
1.717(7)
1.716(15)
1.728(5)
1.716
1.122(12)
1.197(9)
1.168(1)
1.196
149.2(6)
143.8
143.4(9)
142.5
2.001(3)
23
24
21
t.w.
Fe(TpivPP)(NO)f
<Fe(Por)(NO)>g
calc:h FeP(NO)
1.981(26)
2.008(3)
2.044/2.015i
Six-Coordinate NO
(N-MeIm)Fe(TPP)(NO)e
(Py)Fe(TpivPP)(NO)f
<(L)Fe(Por)(NO)>g
calc:h (ImH)FeP(NO)
1.750(2)
1.742(5)
1.752(4)
1.749
1.182(3)
1.194(9)
1.174(6)
1.199
137.7(2)
133.4(5)
138.5(11)
139.3
2.173(2)
2.260(5)
2.008(13)
2.009(2)
2.007
21
21
21
t.w.
2.236
2.046/2.026i
a L, trans axial ligands. b Np, pyrrole N. c Deut, deuteroporphyrin IX; THF, tetrahydrofuran (solvent). d B3LYP functional, VTZ basis set. e N-MeIm,
N-methylimidazole; TPP, tetraphenylporphyrin. f TpivPP, tetrakis(o-pivalamidophenyl)porphyrin. g “Canonical” values reported by Scheidt. h BLYP functional,
VTZ basis set. i Fe-Np bonds directed away from/toward the FeNO bending (Figure 6).
those reported here, apparently because of a different basis set.)
The spin contamination problem is due to the known tendency
of hybrid functionals to exaggerate the stability of high-spin
states.28 This problem is discussed by Ghosh29 in the context
of iron porphyrins.
DFT-Computed Spectra and Backbonding Correlations.
Vibrational modes were computed via DFT, and salient results
are listed in Table 4. Discrepancies between experimental and
DFT-computed frequencies are often adjusted by applying scale
factors to various classes of vibrational coordinates.30 Standard
values are available for bonds in organic molecules, but not for
metal-ligand bonds. The frequencies in Table 4 are unscaled
and show substantial deviations from experimental values, but
the overall pattern is nevertheless instructive.
For the CO adducts, νCO is overestimated by ∼150 cm-1
,
)
while νFeC is overestimated for the 5-c adduct (by 35 cm-1
but underestimated for the 6-c adduct (by 20 cm-1). A
substantial decrease in νFeC on binding ImH to the 5-c adduct
is observed and computed, consistent with the Fe-C bond
lengthening; simultaneously, νCO increases slightly (observed
and computed), consistent with a very small (0.002 Å) shorten-
ing of the CO bond (Table 3). There is also a Fe-C-O bending
mode, computed at 509 and ∼570 cm-1 for the 5-c and 6-c
adducts (there are two slightly different values for the 6-c adduct,
corresponding to bending in and out of the plane defined by
the ImH ligand). This mode is not normally enhanced in RR
spectra for reasons of symmetry, but it has been detected in
some heme protein CO adducts (where the symmetry can be
lowered by steric or electrostatic forces) at ∼570 cm-1 31 in
exact agreement with the computed 6-c value.
Figure 6. Geometry comparison between experimental (see Table 3;
Scheidt’s “canonical” values are used for Fe(II)NO adducts) and compu-
tational (italics) results of Fe(II)P(XO) and (ImH)Fe(II)P(XO).
NO adducts. In particular, the calculation fails to capture the
NO trans effect, as reported earlier.2 The computed Fe-NO
extension on binding ImH is too large, and the Fe-ImH bond
is too short (Table S1 in Supporting Information). This
deficiency can be traced to spin contamination (a problem also
noted by Sage and co-workers25), due to low-lying high-spin
excited states for the NO adducts26 (but not for the CO adducts).
The computed value of S2 is 0.84 using B3LYP, and even higher
for other hybrid functionals (Table S1), whereas the expected
value for one unpaired electron is 0.75. In contrast, non-hybrid
functionals all gave S2 ) 0.77. Among those tried (Table S1),
BLYP gave the best agreement with experimental structure
parameters, and also with vibrational frequencies (next section).
(However, we note that Negrerie et al.27 obtained BLYP-
computed FeP(NO) bond distances somewhat different from
Symmetry also prevents mixing of the Fe-X stretching and
Fe-X-O bending coordinates when the Fe-X-O unit is linear,
(28) Reiher, M.; Salomon, O.; Hess, B. A. Theor. Chem. Acc. 2001, 107, 48-
55.
(29) Ghosh, A. J. Biol. Inorg. Chem. 2006, 11, 712-724.
(30) (a) Rauhut, G.; Pulay, P. J. Am. Chem. Soc. 1995, 117, 4167-4172. (b)
Rauhut, G.; Jarzecki, A. A.; Pulay, P. J. Comput. Chem. 1997, 18, 489-
500.
(31) (a) Tsubaki, M.; Srivastava, R. B.; Yu, N.-T. Biochemistry 1982, 21, 1132-
1140. (b) Vogel, K. M.; Spiro, T. G.; Shelver, D.; Thorsteinsson, M. V.;
Roberts, G. P. Biochemistry 1999, 38, 2679-2687. (c) Fan, B.; Gupta, G.;
Danzier, R. S.; Friedman, J.; Rouessau, D. L. Biochemistry 1998, 37, 1178-
1184. (d) Vogel, K. M.; Hu, S.; Spiro, T. G.; Dierks, E. A.; Yu, A. E.;
Burstyn, J. N. J. Biol. Inorg. Chem. 1999, 4, 804-813.
(25) Leu, B. M.; Zgierski, M. Z.; Wyllie, G. R. A.; Scheidt, W. R.; Sturhahn,
W.; Alp, E. E.; Durbin, S. M.; Sage, J. T. J. Am. Chem. Soc. 2004, 126,
4211-4227.
(26) Rovira, C.; Kunc, K.; Hutter, J.; Ballone, P.; Parrinello, M. J. Phys. Chem.
A 1997, 101, 8914-8925.
(27) Negrerie, M.; Kruglik, S. G.; Lambry, J. C.; Vos, M. H.; Martin, J. L.;
Franzen, S. J. Biol. Chem. 2006, 281, 10389-10398.
9
J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006 16839

A R T I C L E S
Ibrahim et al.
Table 4. Experimental and Computed Frequencies and Main Potential Energy Contributions to FeXO Modes of Fe(II)NO and Fe(II)CO
Porphyrins
calculated
experimental
CO
NO
54
/
14/
Fe(TPP)(CO)
Fe(TPP)(NO)
∆
57Fe
∆
15N
1
1
1
1
1
1
(cm-
)
(cm-
)
ν
(cm-
)
PEDa (%)
ν
(cm-
)
PED (%)
(cm-
)
(cm-
)
Five-Coordinate
δFeXO
νFe-XO
νX-O
510
561
2101
65
88
97
425
599
1686
32%
95%
96%
3.7
4.5
0.1
3.0
10.4
32.0
525b
524c
1951b
1678c
Six-Coordinate
51/52
δFeXO
νFe-XO
νX-O
451d
582e
568/573
474
472
602
31% δFeNO
5.8
2.6
0
2.9
14.8
31.2
29% νFe-NO
66% νFe-NO
21% δFeNO
97%
495e
84
97
1962e
1623e
2111
1669
a Percentage contribution of the indicated internal coordinate to the potential energy of the computed mode. b Reference 3. c Reference 2. d Reference 10.
e This work.
normal-coordinate analysis”, in which selected force constants
initially computed via DFT are refined empirically to fit the
spectral data. They found the predominant contribution to be
Fe-N stretching for the 444 cm-1 mode, and Fe-N-O bending
for a 530 cm-1 mode, which they identified in the RR spectrum
of (N-MeIm)FeTPP(NO). However, 530 cm-1 is the isotope-
sensitive RR band for 5-c FeTPP(NO); in (N-MeIm)FeTPP-
(NO), the band shifts up to 582 cm-1 (Table 1). We infer that
the RR laser induced ligand dissociation in Praneeth et al.’s
sample, even though it was contained in a KBr pellet.
Consequently, the force constant refinement, and therefore the
potential energy distribution, could not have been correct.
Figure 7. Eigenvectors of Fe(II)NO “bending” and “stretching” modes
Sage and co-workers10 also proposed reassigning the 556
(computed frequencies). The stretching and bending internal coordinates
cm-1 mode in MbNO (the frequency is lower than in the protein-
are substantially mixed (see Results and Discussion).
free 6-c adducts, as discussed below) from νFeN to δFeNO,
because its NRVS band has a smaller Fe displacement than does
another NRVS band at 451 cm-1. In addition, NRVS on an
oriented crystal of (N-MeIm)FeTPP(NO) revealed significant
Fe motion perpendicular to the porphyrin plane for the lower
frequency 440 cm-1 mode, but not for the higher frequency
540 cm-1 mode. Qualitatively, these observations are not
inconsistent with our computed mode compositions. The eigen-
vectors (Figure 7) show a larger Fe displacement (including
the out-of-plane component) for the 472 cm-1 than the 601 cm-1
band, consistent with a larger computed 54/57Fe isotope shift for
the former (Table 4).
as it is in the CO adducts. The potential energy entries in Table
4 confirm that the νFeC mode is essentially pure Fe-C
stretching. However, the Fe-C-O bending contribution to
δFeCO is only 50-65%, because of mixing with the Fe-C
tilting coordinate.32
When the Fe-X-O unit is bent, there is mixing between
Fe-X stretching and Fe-X-O bending coordinates, producing
two modes of mixed composition; the mixing also drives the
mode frequencies apart. This phenomenon has been analyzed
with empirical force fields in the context of protein-induced
distortions of the Fe-C-O unit,33 and the natural bending of
the Fe-N-O unit.34 Does this mixing vitiate the labeling of
one of these modes as “νFeN”? Table 4 indicates that the ∼600
cm-1 mode, observed at 524 and 582 cm-1 in 5-c and 6-c NO
adduct and identified via its large 15NO shift, is indeed νFeN.
The Fe-N stretching contribution is 95% for the 5-c adduct
and 66% for the 6-c adduct. The mode having the highest Fe-
N-O bending contribution is computed to be at 424 and 471
However, the mode compositions in Table 4 may not be
entirely correct, because there are discrepancies between the
computed and experimental frequencies. A point of particular
interest is the difficulty in reproducing the frequency upshift of
νFeN, from 524 to 582 cm-1 upon binding N-MeIm to νFeN.
As pointed out by Sage,10 a decrease in νFeN might have been
expected because of the ligand competition, and indeed the
Fe-N bond distance does lengthen, as noted above, although
not by as much as does the Fe-C distance in the CO adduct
(0.03 vs 0.06 Å). Consistent with this lengthening, the Fe-N
force constant diminishes (3.61 mdyn/Å in FeP(NO) vs 3.21
mdyn/Å in (ImH)FeP(NO)), although, again, not by as much
as does the Fe-C force constant (3.51 mdyn/Å in FeP(CO) vs
2.44 mdyn/Å in (ImH)FeP(CO)). Countering the reduction in
force constant, however, is a reduction in effective mass, from
17.9 amu for the 599 cm-1 mode of FeP(NO) to 14.7 amu for
the 602 cm-1 mode of (ImH)FeP(NO), as a result of the νFeN/
cm-1 for 5-c and 6-c adducts, and to have a small (3 cm-1
)
15NO shift. The eigenvectors for these modes are shown in
Figure 7.
Praneeth18 et al. identified bands at 371 and 444 cm-1 in IR
spectra of FeTPP(NO) and (N-MeIm)FeTPP(NO) as 15N18O-
sensitive. They assigned the first of these to δFeNO, but the
second to νFeN, based on a “quantum-chemistry-centered
(32) Ghosh, A.; Bocian, D. F. J. Phys. Chem. 1996, 100, 6363-6367.
(33) Li, X. Y.; Spiro, T. G. J. Am. Chem. Soc. 1988, 110, 6024-6033.
(34) Hu, S. Z.; Kincaid, J. R. J. Am. Chem. Soc. 1991, 113, 9760-9766.
9
16840 J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006

NO and CO Vibrations in Heme Adducts
A R T I C L E S
Figure 9. Three highest HOMOs with major contribution from the FeNO
moiety of Fe(II)P(NO) and (ImH)Fe(II)P(NO).
Ghosh’s studies35). In the two highest occupied orbitals, the NO
antibonding π* orbital in the bending plane (xz) interacts
2
strongly with the dz orbital on the one hand and the dxz orbital
on the other. The third HOMO involves a classical backbonding
interaction, between the out-of-plane π* orbital and dyz. In the
CO adducts, the two highest orbitals are both of this backbond-
ing character, while the third involves porphyrin orbitals only
(Figure S1). Thus, the backbonding correlations for CO adducts
result from a straightforward dπ-π* interaction, which is
influenced by inductive substituent effects. For NO adducts,
Figure 8. νFeX/νXO anti-correlation computed via DFT (B3LYP for CO,
BLYP for NO) for FeP-Y(XO) (0 CO, 9 ΝÃ) and (ImΗ)FeP-Y(XO)
(O CO, b NO); the Y substituents and their attachment to Câ or Cµ atoms
of the porphine are indicated. Lines are drawn with the indicated least-
squares slopes. “Bent” refers to (ImH)FeP-Y(NO) (labeled as 2) calcula-
tions in which the FeNO angle was constrained at 129°.
2
however, the complex interplay between dxz and dz may be
influenced by the porphyrin π orbitals (which are no longer
2
orthogonal to dz because of its tilt (Figure 9)), and this influence
may vary for different substituents when they are directly on
the ring. We note that similar complexities were found by
Rodgers and co-workers in computing backbonding correlations
for Fe(III)NO adducts.36
δFeNO coordinate mixing. The mode frequency does go up,
but the 3 cm-1 computed elevation is far less than the 58 cm-1
observed elevation. Thus, the effective mass is still overesti-
mated, suggesting that coordinate mixing would be greater if
the mode were properly calculated.
Protein Effects and the FeXO Angle. In Figure 10, we plot
νFeX/νXO data from the literature for both CO and NO adducts
of several heme proteins (see Table 5). The lines are those
obtained experimentally from the protein-free TPP-Y porphy-
rins (Figure 5). The 6-c protein-free CO line describes the
protein-CO adducts quite well, as expected from many previous
studies.4 There are certain, well-understood exceptions. Thus,
cytochrome P450 falls below the line because imidazole is
replaced as axial ligand by thiolate, a stronger donor, which
further weakens the Fe-CO bond. On the other hand, the
positive deviation for cytochrome oxidase is believed to result
from compression of the Fe-CO bond by a Cu+ ion, which
lies close to the distal end of the bound CO in the binuclear
heme-Cu site.4 The remaining proteins fall close to the line at
positions that reflect the backbonding influence of distal groups.
Thus, wild-type myoglobin (WT Mb) lies high on the line (low
νCO, high νFeC) thanks to the polar effect of the distal histidine
(H64), which donates a weak H-bond to the bound CO,
enhancing the extent of backbonding. For the Mb variants H64L
and H64I, in which the distal histidine is replaced by the
hydrophobic residues leucine and isoleucine, this effect is absent,
and the points shift down the backbonding line. Other proteins
represented in Figure 10, principally representing heme sensor
In the face of this uncertainty about the exact mode composi-
tion, we propose to be pragmatic, and to simply assign “νFeN”
to the mode near 600 cm-1 that experiences a 15-20 cm-1
14/15N isotope shift (Tables 1 and 4). This is the mode that shows
a very good backbonding correlation, experimentally. (Whether
the ∼450 cm-1 mode also correlates with νNO is uncertain,
because this band is very weak in the RR spectra of the NO
adducts.)
We also examined the backbonding correlations computa-
tionally, by attaching electron-withdrawing or -donating sub-
stituents, Y, to the porphine ring, either at the pyrrole ring Câ
outer atoms or at the Cm atoms that bridge the pyrrole rings.
The CO adducts gave quite good backbonding correlations
(Figure 8), with slopes that were gratifyingly close to the
experimental values (Table 2) for both 5- and 6-c adducts.
(These slopes are considerably improved from those reported
in a previous study,2 due to the use of a larger basis set, VTZ
instead of VDZ.) Similar calculations for the NO adducts were
less successful (Figure 8). The 6-c adducts did correlate linearly,
but the slope was much lower than the experimental slope (Table
2), while the 5-c adducts showed a weak correlation, with
considerable scatter.
(35) Ghosh, A. Acc. Chem. Res. 2005, 38, 943-954.
(36) (a) Linder, D. P.; Rodgers, K. R.; Banister, J.; Wyllie, G. R. A.; Ellison,
M. K.; Scheidt, W. R. J. Am. Chem. Soc. 2004, 126, 14136-14148. (b)
Linder, D. P.; Rodgers, K. R. Inorg. Chem. 2005, 44, 1367-1380.
(37) Andrew, C. R.; George, S. J.; Lawson, D. M.; Eady, R. R. Biochemistry
2002, 41, 2353-2360.
Thus, the backbonding correlations are satisfactorily modeled
for CO but not NO by employing substituents directly attached
to the porphyrin ring. The experimental data are for substituents
located on the phenyl group of TPP, at one removed from the
porphyrin itself. The effect of this difference cannot be evaluated
computationally because the TPP-Y adduct has too many atoms
for practicable DFT calculations with a series of substituents.
(38) Andrew, C. R.; Kemper, L. J.; Busche, T. L.; Tiwari, A. M.; Kecskes, M.
C.; Stafford, J. M.; Croft, L. C.; Lu, S.; Mo¨enne-Loccoz, P.; Huston, W.;
Moir, J. W. B.; Eady, R. R. Biochemistry 2005, 44, 8664-8672.
(39) Tomita, T.; Gonzalez, G.; Chang, A. L.; Ikeda-Saito, M.; Gilles-Gonzalez,
M.-A. Biochemistry 2002, 41, 4819-4826.
(40) Karow, D. S.; Pan, D.; Tran, R.; Pellicena, P.; Presley, A.; Mathies, R. A.;
Marletta, M. A. Biochemistry 2004, 43, 10203-10211.
We speculate that the unsatisfactory modeling of the NO
correlations stems from the complex orbital structure of the NO
adducts, illustrated in Figure 9 (and discussed more fully in
(41) Pinakoulaki, E.; Ohta, T.; Soulimane, T.; Kitagawa, T.; Varotsis, C. J. Biol.
Chem. 2004, 279, 22791-22794.
(42) Pinakoulaki, E.; Ohta, T.; Soulimane, T.; Kitagawa, T.; Varotsis, C. J. Am.
Chem. Soc. 2005, 127, 15161-15167.
9
J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006 16841

A R T I C L E S
Ibrahim et al.
Figure 10. νFeX/νXO plot for available data from heme protein Fe(II)XO adducts (see Table 5). The “b” are H64L and H64I variants of Mb. “]” are
proteins with distal histidine (Ngb, Cgb, and WT-Mb) or, in the case of CcO, with a distal CuB center. The “9” is cyt P450cam bound to adamantanone.
Cytchrome P450 has a cysteine axial ligand; the other proteins all have histidine ligands. The backbonding correlations (solid lines) are for 6-c protein free
adducts (from Figure 5).
Table 5. Vibrational Frequencies for 6-c Fe(II)-XO Heme
line, presumably because of a compressive effect, instead
displaces the NO adduct below the 6-c NO line, to a position
close to WT Mb. The common feature of all four deviant points
is the presence of a positively polar entity, histidine or Cu+, on
the distal side of the bound NO.
We considered the possibility that negative deviations from
the 6-c NO adduct correlation might result from weakening or
breaking of the axial imidazole bond, thereby moving the point
toward the 5-c line (Figure 10). This bond is weak and is easily
stretched2,46 and broken. For example, lowering the pH of
MbNO to 4 is sufficient to break the bond protonate the proximal
histidine.47 However, there is no reason to link axial bond
weakening with distal positive polarity. This issue has been
checked computationally by Tangen et al.,11c who found that
the axial imidazole bond is actually strengthened by distal
H-bond donors.
Proteins
ν
(Fe
−
(cm-
N)
ν
(Fe
−
(cm-
C)
ν
(N
−
(cm-
O)
ν
(C
−
(cm-
O)
1
1
1
1
proteins
)
)
)
)
ref
AXCP
RCCP
579
569
568
560
563
567
553
539
569
573
552
563
558
554
491
494
497
493
487
494
490
507
492
494
508
490
490
474
1624
1625
1634
1637
1632
1639
1655
1620
1604
1600
1612
1635
1638
1591
1966
1966
1967
1973
1969
1972
1989
1973
1971
1972
1941
1965
1968
1955
37
38
39
39
39
39
40
41, 42
43
43
9
BjFixLH
AxPDEA1H
EcDosH
MtDosH
TtTar4H
CcO (T. t.)
Cgb
Ngb
WT-Mb
H64L-Mb
H64I-Mb
P450cam
9
9
34, 44, 45
Earlier we suggested9 that the anomalous νFeN/νNO values
in WT Mb might result from further reduction of the FeNO
angle from its unconstrained value, induced by steric and/or
electrostatic interaction with the distal histidine side chain. Trial
DFT calculations, using B3LYP, produced large reductions in
both νFeN and νNO when the FeNO angle was constrained to
be 10° smaller than its computed equilibrium value, 142°. We
repeated this calculation using BLYP, which gives a better
account of the ground-state structure, as discussed above, and
obtained essentially the same result (Table 6). Reducing the
FeNO angle by 10°, from the equilibrium value of 139°, reduced
νFeN by 26 cm-1, and νNO by 35 cm-1 (the previous
proteins, cluster in the same region, and presumably have
hydrophobic binding pockets.
The NO data are more scattered, but the hydrophobic Mb
variants and the heme sensor proteins nevertheless fall close to
the 6-c protein-free line. Thus, 6-c NO adducts are not
necessarily different in a protein than outside a protein.
Cytochrome P450 falls well below the line, as it does for the
CO adducts, and presumably for the same reason, that the axial
ligand is thiolate, a strong donor.
However, there are striking deviations from the 6-c protein-
free line for some NO adducts, even though the axial ligand
remains imidazole. One of these is WT Mb, which falls far
below the line. Similarly positioned are neuroglobin and
cytoglobin; these share with WT Mb the presence of a distal
histidine residue.43 Also anomalous is cytochrome oxidase. The
distal Cu+ ion that displaces the CO adduct above the 6-c CO
(44) Nagano, S.; Shimada, H.; Tarumi, A.; Hishiki, T.; Kimata-Ariga, Y.; Egawa,
T.; Suematsu, M.; Park, S. Y.; Adachi, S.; Shiro, Y.; Ishimura, Y.
Biochemistry 2003, 49, 14507-14.
(45) Unno, M.; Christian, J. F.; Sjodin, T.; Benson, D. E.; Macdonald, I. D.;
Sligar, S. G.; Champion, P. M. J. Biol. Chem. 2002, 277, 2547-2553.
(46) (a) Mare´chal, J.; Barea, G.; Maseras, F.; Lledo´s, A.; Mouawad, L.; Pe´rahia,
D. J. Comput. Chem. 2000, 21, 282-294. (b) Patchkovskii, S.; Ziegler, T.
Inorg. Chem. 2000, 39, 5354-5364.
(43) Sawai, H.; Makino, M.; Mizutani, Y.; Ohta, T.; Sugimoto, H.; Uno, T.;
Kawada, N.; Yoshizato, K.; Kitagawa, T.; Shiro, Y. Biochemistry 2005,
44, 13257-13265.
(47) Duprat, A. F.; Traylor, T. G.; Wu, G. Z.; Coletta, M.; Sharma, V. S.; Walda,
K. N.; Magde, D. Biochemistry 1995, 34, 2634-2644.
9
16842 J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006

NO and CO Vibrations in Heme Adducts
A R T I C L E S
Table 6. Angle Dependence of (ImH)FeIIP(NO): Frequencies
(cm-1), Distances (Å), Energies (kcal/mol), and Atomic Charges
(z)
angle (deg)
129
139
149
νFeN
νNO
δFeNO
d(Fe-N)
d(N-O)
d(Fe-Im)
∆E
576
1634
467
1.775
1.202
2.230
1.42
-0.019
-0.178
602
1669
471
1.749
1.199
2.236
0
600
1697
474
1.734
1.196
2.247
1.23
+0.007
-0.212
zN
zO
-0.005
-0.195
Figure 12. νFeX/νXO plot for the indicated variants of Mb: WT (stars),
H64X (circles), L29F (circles), V68X (squares), and F46X (diamonds). The
open symbols represent FeIINO adducts; closed symbols are FeIICO adducts.
The solid lines are the backbonding correlations for 6-c protein free XO
adducts (from Figure 5). The inset shows the positions of mutated distal
residues in the Mb binding pocket.49
Figure 11. Valence isomers of LFeP(NO). Although formal charges on
NO are “+” for (I) and “-” for (II), these are compensated by greater
backbonding in (I). However, negative charge builds up on N in (III) and
is stabilized by appropriately oriented H-bonds, as in MbNO.
In both crystallographic studies, the samples were prepared by
nitrosylation of aquo-metMb crystals, so it is conceivable that
sample history somehow influenced the structures.60 Reinforcing
this possibility is early EPR evidence indicating large changes
in the FeNO angle with temperature.51,52
calculation predicted 41 an 74 cm-1 reductions9). In contrast,
increasing the FeNO angle by 10° produced a smaller νNO
increase (28 cm-1) and essentially no change in νFeN. The
difference reflects both electronic and kinematic effects, as
discussed previously.9 Reducing the FeNO angle lengthens the
Fe-NO bond and contracts the Fe-ImH bond slightly, while
increasing the angle has the opposite effect (Table 6). These
trends are consistent with expected changes in resonance
structure (Figure 11). Opening the angle favors the linear Fe-
(I)(NO+) structure, in which the NO is triple-bonded. Its
In any event, the vibrational data support the hypothesis that
properly oriented distal groups with positive polarity can produce
modest closure of the FeNO angle, with large effects on νFeN
and νNO. If this new angle were held constant, then a new
backbonding correlation would presumably apply. We checked
this expectation by recalculating frequencies for the 6-c (ImH)-
FeP-Y(NO) series with the FeNO angle constrained to be 129°;
the resulting data describe a backbonding line, which is shifted
down but is parallel to the line for the equilibrium angle (Figure
8).
With these considerations in mind, we reexamined the data9
on a series of Mb variants with specific residue replacements
in the distal pocket (Figure 12, see inset for the arrangement of
distal groups). The CO adducts all fall on the protein-free 6-c
backbonding correlation, at positions reflecting the control of
backbonding by electrostatic effects. As mentioned above,
hydrophobic replacements of His64 produce points at the low
end of the line, due the loss of H-bonding. However, the H64Q
replacement moves the point only part way down the line,
because the glutamine side chain retains some H-bond donor
ability. At the same time, introduction of a bulky phenylalanine
side chain for Leu29 at the back of the pocket reinforces the
H-bond, either from histidine (L29F variant) or from glutamine
(L29F/H64Q variant), moving both points up the line. On the
other hand, the orientation of His64 in WT Mb is maintained
by the phenyl ring of Phe46, and its replacement by valine
allows the His64 side chain to move away from the CO, leaving
2
antibonding electron is transferred to the Fe dz orbital, lengthen-
ing the Fe-ImH bond. Closing the angle favors the Fe(III)(NO-)
structure, in which two electrons occupy the NO π* orbital,
-
2
producing double-bonded NO , and the dz orbital is emptied,
shortening the Fe-ImH bond. The energies required to increase
or decrease the FeNO angle by 10° are small, 1.4 and 1.2 kcal/
mol, reflecting the electronic flexibility of the FeNO system.
We note that the observed differences between WT Mb and
the H64L variant, which lies near the 6-c adduct line, are 11
cm-1 for νFeN and 23 cm-1 for νNO, roughly one-half the
predicted effects for a 10° closing of the FeNO angle. Thus,
small angle changes can produce the large observed deviations
from the backbonding correlation.
Structural data on the FeNO angle in MbNO are ambiguous.
There are two published X-ray crystal structures of MbNO, one
of which found a surprisingly acute angle, 112°,48 while the
other found a normal angle, 147°,49 as did a frozen solution
EXAFS study.50 The two crystal structures were both at medium
resolution, 1.7 and 1.9 Å, and the discrepancy between them is
a puzzle, especially as there seems to be no significant difference
in the disposition of the distal histidine, or of other distal groups.
(50) Rich, A. M.; Armstrong, R. S.; Ellis, P. J.; Lay, P. A. J. Am. Chem. Soc.
1998, 120, 10827-10836.
(51) Chien, J. C. W. J. Chem. Phys. 1969, 51, 4220-4227.
(52) Hori, H.; Ikeda-Saito, M.; Yonetani, T. J. Biol. Chem. 1981, 256, 7849-
7855.
(48) Brucker, E. A.; Olson, J. S.; Ikeda-Saito, M.; Phillips, G. N. Proteins 1998,
30, 352-356.
(49) Copeland, D. M.; West, A. H.; Richter-Addo, G. B. Proteins 2003, 53,
182-192.
9
J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006 16843

A R T I C L E S
Ibrahim et al.
a hydrophobic environment;53,54 the point for F46V is at the
bottom of the line. Finally, substitution of asparagine for Val68,
which is in direct contact with the bound CO (and is responsible
for the small (8°) deviation from FeCO linearity55), introduces
a strong H-bond; V68N is at the top of the line. Thus, the
strength of the H-bond interaction is directly reflected by the
νFeC/νCO position on the backbonding correlation.
This is definitely not the pattern seen for NO adducts of the
same variants, for which we notice instead that the distal
H-bonds induce a displacement from the 6-c NO line, which
increases with the magnitude of the interaction. The largest
displacement is seen for V68N, which shows the largest
backbonding augmentation for the CO adduct. The substantial
displacement of WT Mb is increased when Leu29 is substituted
by phenylalanine, augmenting the His64 H-bond. On the other
hand, the displacement is diminished when His64 is replaced
by glutamine, the weaker H-bonder, and essentially disappears
for the H64L and H64I hydrophobic replacements. Likewise,
the WT displacement diminishes successively when the but-
tressing effect of Phe46 is relaxed by replacement with smaller
side chains from leucine, valine, or alanine.
by backbonding, which is greater for Fe(I)NO+ (having two
dπ-π* interactions) than for Fe(III)NO- (having one dπ-π*
interaction). Thus, increased FeNO bending would be favored
if a distal H-bond donor is oriented toward the N atom of the
bound NO (Figure 11). This is in fact the orientation of the
distal histidine in MbNO (Figure 12, inset). In contrast, a distal
tyrosine H-bond donor in the TtTar4H domain points toward
the O atom (judging from the crystal structure of the O2
adduct56), and the νFeN/νNO point for this protein does fall on
the 6-c adduct line (Figure 10). We have carried out preliminary
DFT calculations, which support the idea that H-bond donors
affect the FeNO angle if they point at the N, but not the O
atom. This interaction also explains why the CO and NO adducts
are so different in their vibrational responses to H-bond effects
(Figure 11). H-bond donation simply polarizes the FeCO
orbitals, without an inherent effect on the geometry. Because
the FeCO unit lacks the antibonding electron that shifts back
and forth in FeNO with changing angle, the FeCO geometry is
indifferent to polarization. Conversely, modest changes in the
FeCO angle, which can result from steric effects,55 produce
small changes in νFeC and νCO.57
These considerations suggest how the electronic flexibility
of the FeNO unit adds a dimension of complexity to the
interpretation of structure and spectra. The spread of points in
both directions of the νFeN/νNO plots makes NO a highly
sensitive probe of protein influences, more so than CO. The
data reflect not only the presence, but also the directionality of
distal polar groups. It is possible that the N-specific H-bond
effect also accounts for the altered rhombicity seen in temper-
ature-dependent EPR and ENDOR of various heme protein NO
adducts, and characterized as type I and II.58 A distal histidine
signal has been tentatively identified in type II spectra,59 which
may be associated with H-bond anomalies seen in the νFeN/
νNO vibrational data.
This pattern suggests that the distal H-bonds close down the
FeNO angle and that the stronger is the interaction the smaller
is the angle. As seen from our model calculations, the extent of
angle change need not be large to produce the observed
frequency changes, but the direction of the effect is clear. At
the same time, however, H-bonding is also expected to increase
νFeN while decreasing νNO by increasing the backbonding
interaction, just as in the CO adducts. This can explain why
most of the observed displacements from the line mainly involve
νNO, with little variation in νFeN. (The small νFeN variation
is what led Boxer to infer that backbonding is unimportant in
NO adducts.8) If H-bonding decreases the FeNO angle, the
accompanying increase in backbonding would reinforce the
angle-induced diminution of νNO, but would counteract it in
the case of νFeN, producing the observed horizontal spread for
most of the points.
Conclusions
Although the bent FeNO structure of heme-NO adducts
induces mixing of Fe-N stretching and Fe-N-O bending
coordinates, a 15/14NO-sensitive vibrational band in the 500-
600 cm-1 region can still be identified as “νFeN”. Plots of this
frequency against νNO produce negative linear correlations for
both 5- and 6-coordinate adducts of Fe(II)TPP-Y adducts with
variably electron-donating Y substituents, quite similar to νFeC/
νCO correlations of the corresponding CO adducts.
However, CO and NO adducts respond very differently to
protein influences, especially to distal H-bond donors or positive
charges. While the νFeC/νCO points move up and down the
backbonding correlation, in proportion to the strength of the
polar interaction, the νFeN/νNO points deviate strongly from
the correlation, in the same order of interaction strength. These
However, some of the variants show appreciable changes in
νFeN, suggesting backbonding changes independent of angle
changes. Intriguingly, the points for F46V, H64Q, L29F/H64Q,
and V68F are ranged along a line parallel to the 6-c protein
free line. This is the behavior expected for variable backbonding
at constant FeNO angle, as confirmed by our model calculations
on (ImH)FeP-Y(NO) with constrained FeNO angle (Figure 8).
Why should H-bonding affect the FeNO angle? We suggest
that the answer lies in the interaction of geometry with resonance
structure, as mentioned above. Smaller angles are associated
with the Fe(III)NO- resonance structure, and this structure
should be favored by H-bond donors. However, the orientation
of the H-bond donor is critical because the FeNO angle affects
the charge differently on the N and O atoms. The DFT
calculation (Table 6) shows negative charge increasing on N
but decreasing on O as the FeNO angle decreases. The total
charge on NO changes very little. This is because the formal
charges in the valence isomers (Figure 11) are compensated
(56) Pellicena, P.; Karow, D. S.; Boon, E. M.; Marletta, M. A.; Kuriyan, J.
Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 12854-12859.
(57) Kozlowski, P. M.; Vogel, K. M.; Zgierski, M. Z.; Spiro, T. G. J. Porphyrins
Phthalocyanines 2001, 5, 312-322.
(58) (a) Hu¨ttermann, J.; Burgard, C.; Kappl, R. J. Chem. Soc., Faraday Trans.
1994, 90, 3077-3087. (b) Flores, M.; Wajnberg, E.; Bemski, G. Biophys.
J. 1997, 73, 3225-3229. (c) Hori, H.; Ikeda-Saito, M.; Yonetani, T. J.
Biol. Chem. 1981, 256, 7849-7855.
(59) Tyryshkin, A. M.; Dikanov, S. A.; Reijerse, E. J.; Burgard, C.; Hu¨ttermann,
J. J. Am. Chem. Soc. 1999, 121, 3396-3406.
(53) Li, T. S.; Quillin, M. L.; Phillips, G. N.; Olson, J. S. Biochemistry 1994,
33, 1433-1446.
(60) Note added in proof. A very recent crystallographic study (Copeland, D.
M.; Soares, A. S.; West, A. H.; Richter-Addo, G. B. J. Inorg. Biochem.
2006, 100, 1413-1425) at 1.3 Å resolution reveals alternatiVe structures,
dependent on crystal preparation: FeNO ) 144° for metMbIII crystals
treated with nitrite/dithionite, but 120° for MbII crystals treated with NO.
(54) Lai, H. H.; Li, T. S.; Lyons, D. S.; Phillips, G. N.; Olson, J. S.; Gibson, Q.
H. Proteins 1995, 22, 322-339.
(55) De Angelis, F.; Jarzecki, A. A.; Car, R.; Spiro, T. G. J. Phys. Chem. B
2005, 109, 3065-3070.
9
16844 J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006

NO and CO Vibrations in Heme Adducts
A R T I C L E S
deviations are judged to result from diminution of the FeNO
angle due to the polar stabilization of the Fe(III)(NO-) valence
isomer, a stereoelectronic effect that is absent for CO adducts.
This added effect may be useful in probing the determinants of
ligand discrimination and signaling in heme proteins.
Supporting Information Available: Table S1, listing all of
the key structural parameters and S2 values calculated by the
different DFT functionals in this study, with selected structural
information from experiments and other computations for
comparison. Figure S1, showing the three highest MOs of Fe-
(II)P(CO) and (ImH)Fe(II)P(CO). Complete ref 15. This material
is available free of charge via the Internet at http://pubs.acs.org.
Acknowledgment. We thank Dr. Ingar Wasbotten and Dr.
Andrzej Jarzecki for preliminary work on this project, and Prof.
John Olson for helpful discussions. This work was supported
by NIH grant GM 33576.
JA064859D
9
J. AM. CHEM. SOC. VOL. 128, NO. 51, 2006 16845