International Hydrogenases Conference 2004
EPR experiments to elucidate the structure of
the ready and unready states of the [NiFe]
hydrogenase of Desulfovibrio vulgaris Miyazaki F
M. van Gastel, C. Fichtner, F. Neese and W. Lubitz1
Max-Planck-Institut für Bioanorganische Chemie, P.O. Box 101365, D-45413 Mülheim an der Ruhr, Germany
Abstract
Isolation and purification of the [NiFe] hydrogenase of Desulfovibrio vulgaris Miyazaki F under aerobic
conditions leads to a mixture of two states, Ni-A (unready) and Ni-B (ready). The two states are distinguished
by different activation times and different EPR spectra. HYSCORE and ENDOR data and DFT calculations show
that both states have an exchangeable proton, albeit with a different 1 H hyperfine coupling. This proton is
assigned to the bridging ligand between Ni and Fe. For Ni-B, a hydroxo ligand is found. For Ni-A, either a
hydroxo in a different orientation or a hydroperoxo-bridging ligand is present.
Introduction
Molecular hydrogen plays an important role in the metabolism of many microorganisms. Both H2 production and
H2 uptake are catalysed by hydrogenases (see [1] for a
review). In standard [NiFe] hydrogenases, catalysis takes
place at the active site in the large subunit, a [NiFe] bimetallic
centre. Furthermore, three iron-sulphur clusters are present
in the small subunit. Crystal structures of [NiFe] hydrogenases are available for the sulphate-reducing bacteria
Desulfovibrio gigas [2], Desulfovibrio vulgaris Miyazaki F
[3,4], Desulfovibrio fructosovorans [5] and Desulfovibrio
desulfuricans [6]. They all have a conserved co-ordination
geometry of the active site.
The [NiFe] active site consists of one nickel and one iron
atom. The two metals are bridged by two sulphur atoms
provided by two cysteine residues. Two more cysteine ligands
to Ni are present. Three diatomic molecules, one CO and
two CNs, are bound to the iron as shown by FTIR (Fouriertransform infrared) analysis [7–9]. The aerobically isolated
enzyme, referred to ‘as-isolated’, is predominantly present in
the oxidized state and contains a bridging ligand between Ni
and Fe, which is presumably an oxygen species [10]. At least
seven redox states have been assigned to [NiFe] hydrogenases
(see [11] for a review). In the oxidized form, two states coexist;
the Ni-A ‘unready’ and Ni-B ‘ready’ state, which differ in
their activation times, i.e. hours for Ni-A and minutes for
Ni-B. Both states are EPR active and EPR experiments by
many groups have shown that the unpaired electron resides
mainly in a 3dz2 orbital on Ni. When the enzyme is reduced
by two electrons, the so-called Ni-C ‘active’ state is formed,
which is catalytically active. The most reduced state is called
Key words: Desulfovibrio vulgaris Miyazaki F, density functional theory (DFT), electron nuclear
double resonance (ENDOR), EPR, hydrogenase, hyperfine sublevel correlation (HYSCORE).
Abbreviations used: DFT, density functional theory; ENDOR, electron nuclear double resonance;
FTIR, Fourier-transform infrared; HYSCORE, hyperfine sublevel correlation.
1
To whom correspondence should be addressed (email [email protected]).
Ni-R, which is one electron more reduced than Ni-C, EPR
silent and an intermediate in the reaction mechanism of the
enzyme.
For the oxidized states Ni-A and Ni-B, the exact nature
of the additional bridging ligand is still not clear and its
elucidation would be very helpful to understand the oxygen
sensitivity of hydrogenases, which leads to inhibition of the
enzyme. When the enzyme is oxidized to the Ni-A state, it is
inactive and difficult to reactivate. Preliminary X-ray crystallographic data of the Ni-A state indicate the presence
of density between Ni and Fe that may be compatible
with a bridging ligand with two X-ray-detectable atoms,
for example SO, O2 or OOH− (H. Ogata, S. Hirota and
Y. Higuchi, personal communication). The oxygen sensitivity
is a problem that has to be overcome before these enzymes can be used for a possible biotechnological hydrogen
production as a part of clean fuel cells. To address the
question of the identity of the bridging ligand, we performed
FTIR, ENDOR (electron nuclear double resonance) and
HYSCORE (hyperfine sublevel correlation) spectroscopic
measurements and performed DFT (density functional
theory) calculations of the oxidized Ni-A and Ni-B states.
Materials and methods
Sample preparation
Hydrogenase of D. vulgaris Miyazaki F was expressed as
described previously [12]. To prepare the Ni-A state with
maximum yield, as-isolated samples (1.4 mM concentration
in 25 mM Tris buffer, pH 7.4) were first mixed with ascorbate
and phenazine methosulphate to a final concentration of
20 mM and 200 µM respectively. The protein solution was
filled in an EPR tube and gassed with Ar at room temperature
for 1 h. Then the gas was exchanged with CO and the
protein was incubated in CO atmosphere for 1 h at 50◦ C.
To maximize the yield of the Ni-A state, the sample was
C 2005
Biochemical Society
7
8
Biochemical Society Transactions (2005) Volume 33, part 1
Figure 1 Schematic representation of the models used in the DFT calculations and their singly occupied natural orbital
(a) O2− bridge, (b) OH− bridge, first orientation, (c) OH− bridge, second orientation, (d) H2 O bridge, (e) OOH− bridge, first
orientation, (f) OOH− bridge, second orientation. The models consist of Ni, Fe, 2CN, CO, 4SCH2 CH3 (cysteine residues), N2 C4 H6
(histidine residue) and the bridging ligand. Colour code: S, yellow; C, green; H, white; N, mid blue; O, red; metals, light blue.
additionally oxidized in air for 2 min at room temperature.
The yield of the Ni-B state was maximized by reducing the
hydrogenase solution (1.4 mM) with 1 bar of H2 for 6 h
at 37◦ C in an EPR tube. Then the sample was oxidized in
air during incubation on ice for 10 min. For preparation of
the samples in 2 H2 O a buffer exchange was performed. The
hydrogenase solution was diluted (1/200) in 25 mM Tris/
2
H2 O buffer, pD = 7.8 (according to pD = pH + 0.41) and
subsequently concentrated in a Centricon tube (30 kDa;
Millipore). The procedure of dilution and concentration
was repeated 10 times. It should be noted that the samples
were treated according to the above-described procedures
following buffer exchange to 2 H2 O to maximize the amount
of Ni-A or Ni-B respectively.
FTIR spectroscopy
The FTIR experiments of Ni-A and Ni-B hydrogenases were
performed at 30◦ C for Ni-B and 15◦ C for Ni-A on a Bruker
IFS 66v/S FTIR spectrometer using an optically transparent
thin-layer electrochemical cell in which the potential can be
controlled [13].
ENDOR and HYSCORE spectroscopy
ENDOR and HYSCORE measurements were performed
with a Bruker ESP 380E spectrometer and a Bruker dielectric
ring resonator (ESP380-1052 DLQ-H) equipped with an
Oxford helium cryostat CF 935. For ENDOR, a Davies pulse
sequence was used. The length of the π /2 pulse was 56 ns and
the RF pulse had a length of 12 µs. The temperature was
10 K and the microwave frequency 9.72 GHz. In the two
C 2005
Biochemical Society
dimensional-pulse HYSCORE experiment, the pulse sequence of π /2-τ -π /2-t1 -π -t2 -π /2-τ -echo [14,15] was applied
with a duration of 16 and 32 ns for the π /2 and π pulses respectively. The experimental settings were T = 10 K, ν mw =
9.77 GHz, τ = 120 ns. A four-step phase cycle was used
to eliminate contributions from unwanted echoes from the
echo envelope [16]. The baseline of the two-dimensional
modulation pattern was subtracted by a polynomial background function of second order. Subsequently, the time
traces were multiplied with a Hamming function and zerofilled to 1024 × 1024 points. The data were then Fouriertransformed to obtain magnitude contour spectra.
DFT calculations
All DFT calculations were performed with the ORCA
programming package [17]. The B3LYP functional was used
within a spin-unrestricted formalism. For the calculation of
the hyperfine tensors, the large CP basis [18,19] was used for
Ni and Fe. This basis was augmented with two polarization
functions [20] and one additional f function taken from the
turbomole library. The iglo-iii basis set was used for all other
atoms [21]. An effective spin–orbit coupling operator [22–24]
was employed. The model systems of the active site consist
of Ni, Fe, four SCH2 CH3 groups to model the cysteine
residues, the diatomic ligands of iron (two CN− and one
CO), a methyl imidazole group to model His-88 and the
bridging ligand. In this work, the following bridging ligands
were considered: O2− (model a), OH− in two possible
conformations (models b and c), H2 O (model d) and OOH−
in two conformations (models e and f). The geometries of the
models are given in Figure 1. A geometry was considered to
International Hydrogenases Conference 2004
be converged when the change in energy was <10−5 Hartree
(Eh), the average force was <5 × 10−4 Eh/Bohr and the maximum force was <10−3 Eh/Bohr. The convergence criteria
for the self-consistent field part were chosen to be 10−6 Eh
for the change in energy, 10−5 Eh for the change of elements
of the density matrix and 10−6 Eh for the maximum element of the direct inversion of iterative subspace error.
Results and discussion
Figure 2 1 H HYSCORE spectra for Ni-A (left) and Ni-B (right) in
H2 O (a) and in 2 H2 O (b), recorded at the gy canonical orientation
(c) Simulation of the proton signals using hyperfine parameters from DFT
calculation from model b for Ni-A and model c for Ni-B (see Table 1).
(d) 14 N and 2 H region of the HYSCORE spectra in 2 H2 O. The red (Ni-A) and
green (Ni-B) coloured signals correspond to one exchangeable proton
in (a). These lines are missing in the 2 H2 O exchanged sample (b), the
original positions being also indicated in red and green. The signals in
blue boxes in (d) are cross-peaks that belong to the 14 Nε of His-88.
Experimental conditions: ν mw = 9.71 GHz, T = 5 K, τ = 120 ns. For the
preparation of the states, see text.
FTIR spectroscopy
The crystal structure of D. vulgaris Miyazaki F [3] in the oxidized state, a mixture of Ni-A and Ni-B, reported two COs
and one SO ligand attached to the iron. This contradicts the
findings obtained by FTIR spectroscopy of other [NiFe] hydrogenases, in which invariably IR frequencies corresponding
to one CO and two CN ligands are found [7–9]. FTIR
experiments in our laboratory show that the enzyme of D.
vulgaris Miyazaki F also has one CO and two CN ligands
with frequencies of 1956 cm−1 (CO), 2083 cm−1 (CN) and
2093 cm−1 (CN) for Ni-A and 1955 cm−1 (CO), 2081 cm−1
(CN) and 2090 cm−1 (CN) for Ni-B (error ±1 cm−1 ). For a
good description of the electronic structure of the active site
by DFT calculations, it is important to include these ligands
correctly. Therefore the model in our DFT calculations is
based on the X-ray data of D. vulgaris Miyazaki F modified
according to these findings.
HYSCORE and ENDOR spectroscopy
HYSCORE spectra for the Ni-A and Ni-B states in H2 O and
2
H2 O are given in Figures 2(a) and 2(b). The preparation of
both Ni-A and Ni-B, after the buffer exchange to 2 H2 O, is
done by first reducing the enzyme to Ni-C followed by reoxidation (see Materials and methods section). The 1 H region
of the spectra shows signals of exchangeable protons, indicated in red and green for Ni-A and Ni-B respectively (Figure 2a). The ridges of the red and green signals, which are
shifted upwards with respect to the antidiagonal passing
through the 1 H Zeeman frequency (∼14 MHz), appear
different, indicating that they belong to strongly, but differently coupled protons of the [NiFe] site. Simulations are
shown in Figure 2(c) and the hyperfine coupling constants,
hfcs, are given in Table 1.
In Figure 2(d), the 14 N and 2 H region of the spectra
measured in 2 H2 O is presented. Deuterium signals can be
observed, indicated in red and green, which are absent in the
spectra in H2 O (results not shown). The other signals in
these spectra belong to the 14 Nε of His-88, which is involved
in a hydrogen bond to the sulphur of the Ni-co-ordinating
Cys-549. Also the ridges of the 14 N cross-peaks (blue) appear
different for Ni-A and Ni-B, indicating differences in the
electronic structures of these redox states.
Similar observations have been made in single-crystal
ENDOR spectroscopy on oxidized single crystals (results
not shown). For Ni-B, two bands with shifts of 5 and 6 MHz
have been observed. They vary only slightly with the
rotation angle, which indicates
Cys-βCH2 protons with a large
anisotropic hyperfine coupling. A
which is well separated from the
that they stem from
isotropic and a small
third signal is present,
structured matrix peak
C 2005
Biochemical Society
9
10
Biochemical Society Transactions (2005) Volume 33, part 1
Table 1 Calculated (a–f) and experimental (Exp.) hyperfine coupling constants (in MHz) of various models for the active site of Ni-A and
Ni-B hydrogenases in the oxidized Ni-A and Ni-B states
(a) O2−
61 Ni*
Bridging 17 O*
Bridging 1 H
Cys-549 1 H-β 1
Cys-549 1 H-β 2
Cys-549
33 S
His-88 1 Hε
(b) OH−
(c) OH−
(e) OOH−
(d) H2 O
(f) OOH−
Ni-A Exp.
Ni-B Exp.
−45†
−30†
−44†
−30†
aiso
A x
5.9
−27.7
−67.6
−49.0
−73.0
−50.6
−39.5
−41.6
−74.5
−52.6
−108.3
−26.8
A’y
A z
aiso
−6.9
34.6
−40.3
2.9
46.1
−15.9
19.2
31.3
−11.4
−8.3
49.8
−29.7
−16.5
3.1
49.5
−3.9
−18.9
3.7
23.1
0.7
5†
25†
−10.7‡
A x
A y
A z
−168.8
79.9
88.9
−4.1
1.1
2.9
−6.5
1.8
4.7
−5.0
2.1
2.8
−2.7
0.6
2.0
−3.3
0.8
2.5
−10.9
1.4
9.5
−6.9
0.8
6.1
−9.3‡
3.6‡
5.7‡
aiso
A x
4.6
−5.0
−0.8
−5.8
3.9
−4.7
0.0
−4.8
1.7
−2.0
−0.1
−2.6
2.6§
−3.6§
−3.9
−4.3
A y
A z
aiso
−4.2
9.9
9.36
−3.4
8.1
−3.8
8.6
11.29
−1.5
3.6
9.00
−1.1
3.7
12.57
−3.6§
7.2§
−0.12
−4.1
9.1
9.52
−3.1
7.5
13.1
A x
A y
A z
−0.96
−0.87
1.83
−1.50
−0.91
2.40
−1.50
−0.94
2.44
−1.64
−0.76
2.40
−1.46
−0.86
2.32
−1.48
−0.79
2.27
−2.2
−1.7
3.9
aiso
A x
A y
0.47
−1.95
−1.74
7.84
−2.31
−1.41
8.11
−2.34
−1.43
7.14
−2.37
−1.38
7.08
−2.26
−1.39
14.28
−2.36
−1.29
11.2
−1.1
−1.1
A z
aiso
A x
3.68
8.4
−5.0
3.72
37.7
−21.4
3.77
43.1
−21.9
3.76
34.4
−20.9
3.66
37.6
−20.2
3.64
30.2
−21.0
2.2
A y
A z
aiso
−1.0
5.9
−0.17
−14.9
36.3
−0.66
−15.9
37.7
−0.72
−14.9
35.8
−0.61
−13.6
33.6
−0.61
−14.9
35.8
−1.55
A x
A y
−0.91
−0.83
−2.61
−1.81
−2.75
−1.92
−2.31
−1.60
−2.52
−1.70
−2.62
−1.68
A z
1.74
4.42
4.67
3.90
4.21
4.30
14†
17†
*Note that 61 Ni and 17 O have negative nuclear g values (Bruker EPR/ENDOR Frequency Table, 2004).
†From EPR experiments of 61 Ni-labelled D. vulgaris Miyazaki F (A. Goenka et al., unpublished work).
‡From [10].
§Estimates from the HYSCORE spectra.
close to ν H . This signal varies strongly with orientation
and has a frequency shift of approx. 1 MHz at θ = 0◦ and
4 MHz near θ = 100◦ . For this signal, the isotropic and anisotropic hyperfine interactions are of similar magnitude as
those found in the HYSCORE spectra. For Ni-A, the two
almost isotropic signals with shifts of approx. 5 MHz as
observed for Ni-B remain, the anisotropic signal as observed
for Ni-B is not present. The signals in the Ni-A spectra are
also somewhat broader than those for Ni-B, which may be
caused by different relaxation times (M. van Gastel, M. Stein,
M. Brecht, F. Lendzian, R. Bittl, H. Ogata, Y. Higuchi and
W. Lubitz, unpublished work).
DFT calculations
Calculations of several geometry-optimized model systems
for the active site of D. vulgaris Miyazaki F hydrogenase
in the Ni-A and Ni-B oxidation states have been performed
C 2005
Biochemical Society
(Figure 1). Calculated hfcs for all the considered models are
given in Table 1. As a starting point, the X-ray structure
of the oxidized state for D. vulgaris Miyazaki F was used
[3]. Previous calculations indicated that a sulphur-bridging
ligand is not compatible with the X-ray structure, since
the Ni–Fe distance becomes too large {3.2 Å (1 Å = 0.1 nm)
in the calculation [25] and 2.55 Å in the X-ray structure
[3]}. Furthermore, from ENDOR experiments of D. gigas
hydrogenase in an H2 17 O-enriched solvent, the bridging
ligand was found to be an oxygen species [10]. Therefore, we
only considered the X-ray structure for D. vulgaris Miyazaki
F with oxygen-based bridging ligands.
For the OH− -bridging ligand, two binding conformations
are possible that differ in the orientation of the proton/lonepair direction (Figures 1b and 1c). For an OOH− ligand,
even more orientations are possible. The orientation of
the bridging ligands is important when performing DFT
International Hydrogenases Conference 2004
calculations, as the calculated hfcs of the bridging 17 O
and 1 H will depend on the relative position of this ligand
with respect to the spin-carrying 3dz2 orbital on Ni (see
Figure 1).
Comparison of experimental and calculated data (Table 1)
shows that model a with a µ-oxo bridge (O2− ) is not a good
candidate for either Ni-A or Ni-B due to the following
reasons: (i) the model lacks an exchangeable proton with
an hfc of the correct magnitude that has been detected for
both states; (ii) the calculated 61 Ni hyperfine tensor disagrees
with the experimental one, in particular with respect to the
isotropic value; and (iii) the calculated 17 O hyperfine tensor
disagrees with the experimental one (for Ni-A [10]). H2 O
(model d) can also be excluded as a bridging ligand since only
one exchangeable proton has been detected both for Ni-A and
Ni-B by HYSCORE and ENDOR spectroscopy. In models
b and c with OH− as a bridging ligand, both the experimental
and calculated 61 Ni and 17 O hyperfine tensors are of the
correct size. A distinction is possible by the measured hfc
of the exchangeable proton, showing that Ni-B is closer to
model c; furthermore, the angular dependence observed for
this proton in single crystal ENDOR can only be reproduced
by this model. For Ni-A reasonable agreement is found for
the 61 Ni hfc of models b (OH− ) and e (OOH− ) and for the
17
O hfc of models b (OH− ) and f (OOH− ), yielding no
clear answer. For the bridging 1 H, the OH− model b fits better into the experimental data than the OOH− models (e, f)
since the 1 H hfc tensors are too small in the latter. However,
a different geometry of the hydroperoxo ligand placing the
proton somewhat closer to the spin-carrying Ni(3dz 2 ) orbital
could increase the hfc and would thus bring the value closer
to the experimental one. An OOH− -type ligand cannot be
excluded at present for the Ni-A state. Labelling of this ligand
by 17 O and a detection of the respective hfcs by ENDOR
could solve this discrepancy.
The 1 H hfc tensor of the β-CH2 group of Cys-549, which is
carrying a large portion of the spin-density [25], is reproduced
by all models except a, which indicates that the wavefunction
describes satisfactorily the delocalization of spin-density over
the [NiFe] site. The 33 S hfc of this residue is calculated
to be fairly large in models b–f, as expected. A hydrogen
bond formed between the sulphur of Cys-549 and the Nε of
His-88 may also be exchangeable. However, the calculated
1
H hfc tensor (Table 1) is significantly smaller than the
experimentally observed 1 H hfc tensors of the exchangeable
protons in Ni-A and Ni-B. It is therefore concluded that the
respective exchangeable protons indeed belong to the bridging ligand between Ni and Fe and not to the H-bond of the
histidine.
Summary and conclusions
Both Ni-A and Ni-B have one exchangeable proton that is
assigned to the bridging ligand. Comparison between EPR
experiments and DFT calculations shows that this ligand is
OH− for Ni-B. For Ni-A two models exist at present. In
the first one, the bridging ligand is also an OH− but with a
different binding geometry. The second model postulates a
diatomic-type bridging ligand. The latter is supported by
recent X-ray crystallographic data on Ni-A in D. vulgaris
Miyazaki F single crystals. Here, the second atom (oxygen)
is in a position to block the free co-ordination site at the
nickel, leading to inhibition of the hydrogenase similar to
the CO-inhibited enzyme [26].
We thank Dr H. Ogata, Professor S. Hirota and Professor Y. Higuchi
for helpful discussions concerning the structure of Ni-A. This work
was supported by Deutsche Forschungsgemeinschaft (Sfb 498, TP
C2), Fonds der Chemischen Industrie (WL) and by the Max Planck
Society.
References
1 Vignais, P.M., Billoud, B. and Meyer, J. (2001) FEMS Microbiol. Rev. 25,
455–501
2 Volbeda, A., Charon, M.-H., Hatchikian, E.C., Frey, M. and
Fontecilla-Camps, J.C. (1995) Nature (London) 373, 580–587
3 Higuchi, Y., Yagi, T. and Yasuoka, N. (1997) Structure 15, 1671–1680
4 Higuchi, Y., Ogata, H., Miki, K., Yasuoka, N. and Yagi, T. (1999) Structure
7, 549–556
5 Nicolet, Y., De Lacey, A.L., Vernède, X., Fernandez, V.M., Hatchikian, E.C.
and Fontecilla-Camps, J.C. (2001) J. Am. Chem. Soc. 123, 1596–1601
6 Matias, P.M., Soares, C.M., Saraiva, L.M., Coelho, R., Morais, J., LeGall, J.
and Carrando, M.A. (2001) J. Biol. Inorg. Chem. 6, 63–81
7 De Lacey, A.L., Hatchikian, E.C., Volbeda, A., Frey, M., Fontecilla-Camps,
J.C. and Fernandez, V.M. (1997) J. Am. Chem. Soc. 119, 7181–7189
8 George, S.J., Kurkin, S., Thorneley, R.N.F. and Albracht, S.P.J. (2004)
Biochemistry 43, 6808–6819
9 De Lacey, A.L., Fernandez, V.M., Rousset, M., Cavazza, C. and Hatchikian,
E.C. (2003) J. Biol. Inorg. Chem. 8, 129–134
10 Carepo, M., Tierney, D.L., Brondino, C.D., Yang, T.C., Pamplona, A.,
Telser, J., Moura, I., Moura, J.J.G. and Hoffman, B.M. (2002)
J. Am. Chem. Soc. 124, 281–286
11 Albracht, S.P.J. (1994) Biochim. Biophys. Acta 1188, 167–204
12 Yagi, T., Kimura, K., Daidoji, H., Sakai, F., Tamura, S. and Inokuchi, H.
(1976) J. Biochem. 79, 661–671
13 Moss, D., Nabedryk, E., Breton, J.L.J. and Mäntele, W. (1990)
Eur. J. Biochem. 187, 565–572
14 Höfer, P., Grupp, A., Nebenführ, H. and Mehring, M. (1986)
Chem. Phys. Lett. 132, 279–282
15 Shane, J.J., Höfer, P., Reijerse, E.J. and de Boer, E. (1992)
J. Magn. Reson. 88, 241–256
16 Gemperle, C., Aebli, G., Schweiger, A. and Ernst, R.R. (1990)
J. Magn. Reson. 88, 241–256
17 Neese, F. (2003) ORCA – an ab initio, density functional and
semiempirical program package, version 2.2, revision 41, Max Planck
Institut für Bioanorganische Chemie, Mülheim
18 Neese, F. (2002) Theor. Inorg. Chim. Acta 337C, 181–192
19 Weigend, F. and Haser, M. (1997) Theor. Chem. Acc. 97, 331–340
20 Wachters, A.J. (1970) J. Chem. Phys. 52, 1033–1036
21 Kutzelnigg, W., Fleischer, U. and Schindler, M. (1991) in NMR Basic
Principles and Progress, vol. 213, pp. 165–262, Springer-Verlag,
Heidelberg
22 Koseki, S., Schmidt, M.W. and Gordon, M.S. (1992) J. Phys. Chem. 96,
10768–10772
23 Koseki, S., Gordon, M.S., Schmidt, M.W. and Matsunaga, N. (1995)
J. Phys. Chem. 99, 12764–12772
24 Koseki, S., Schmidt, M.W. and Gordon, M.S. (1998) J. Phys. Chem. A 102,
10430–10435
25 Stein, M. and Lubitz, W. (2001) Phys. Chem. Chem. Phys. 3, 2668–2675
26 Ogata, H., Mizogushi, Y., Mizuno, N., Miki, K., Adachi, S., Yasuoka, N.,
Yagi, T., Yamauchi, O., Hirota, S. and Higuchi, Y. (2002) J. Am. Chem. Soc.
124, 11628–11635
Received 22 October 2004
C 2005
Biochemical Society
11