The Fe-only nitrogenase and the Mo nitrogenase
from
Rhodobacter capsulatus
A comparative study on the redox properties of the metal clusters
present in the dinitrogenase components
Stefan Siemann*, Klaus Schneider, Melanie Dro¨ ttboom† and Achim Mu¨ ller
Lehrstuhl fu
¨
r Anorganische Chemie I, Fakulta
¨
tfu
¨
r Chemie der Universita
¨
t Bielefeld, Bielefeld, Germany
The dinitrogenase component proteins of the conventional
Mo nitrogenase (MoFe protein) and of the alternative
Fe-only nitrogenase (FeFe protein) were both isolated and
purified from Rhodobacter capsulatus, redox-titrated
according t o t h e same procedures and subjected t o an EPR
spectroscopic comparison. In the course of an oxidative
titration o f the MoFe protein (Rc1
Mo
) t hree significant
S ¼ 1/2 EPR signals deriving from oxidized states of the
P-cluster were detected: (1) a r hombic signal (g ¼ 2.07, 1.96
and 1 .83), w hich showed a bell-shaped redox curve with
midpoint potentials ( E
m
)of)195 mV (appearance) and
)30 mV (disappearance), (2) a n axial sig nal (g

a shift of the g values to 2.22 and 2.05 and by the appearance
of an additional negative absorption-shaped peak at
g ¼ 1.86. (c) A very narrow rhombic E PR signal at
g ¼ 2.00, 1.98 a nd 1.96 appeared at positiv e redox potentials
(E
m
¼ 80 mV, intensity maximum at 160 mV). Another
novel S ¼ 1/2 signal at g ¼ 1.96, 1.92 and 1 .77 was observed
on further, enzymatic reduction of the d ithionite-reduced
state o f Rc1
Fe
with the dinitrogenase reductase component
(Rc2
Fe
) of t he same enzyme system (turnover conditions in
the presence of N
2
and ATP). When the Rc1
Mo
protein was
treated analogously, neither this Ôturnover signalÕ nor any
other S ¼ 1/2 signal were dete ctable. All Rc1
Fe
-specific EPR
signals detected are discussed and tentatively assigned with
special consideration o f the reference s pectra obtained from
Rc1
Mo
preparations.
Keywords: Fe nitrogenase; FeFe cofactor; FeMo cofactor;

¨
r Chemie, Universita
¨
t Bielefeld, Postfach 100131, 33501
Bielefeld, Germany. Fax: + 49 521 1066003,
Tel.: + 49 521 1 066153, E-mail: a.mue
Abbreviations: nif, nitrogen fixation; vnf, vanadium dependent nitro-
gen fixation; anf, alternative nitrogen fixation; FeMoco, iron–molyb-
denum cofactor; FeFeco, iron–iron cofactor; Rc1
Mo
, MoFe protein of
R. capsulatus;Rc1
Fe
,FeFeproteinofR. capsulatus;Rc2
Mo
,Fepro-
tein of the Mo nitrogenase of R. capsulatus;Rc2
Fe
, Fe protein of the
Fe-only nitrogenase of R. capsulatus; EXAFS, extended X-ray
absorption fine struc t ure.
Enzyme: nitrogenase (EC 1.18.6.1).
*Presen t address: Department o f Chemistry, University of Waterloo,
Waterloo, Ontario, Canada.
Presen t address: Transferstelle Umweltbiotechnology,
Ruhr-Univer sita
¨
t Bochum, 447 80 Bochum, G ermany.
(Received 1 9 September 200 1, revised 2 8 December 200 1, accepted
22 January 2002)

7
MoS
9
/homocitrate) and the P-cluster
(Fe
8
S
7
) have been elucidated [12,13], the specific site(s) of
substrate binding and reduction within the cofactor, how-
ever, still remain a matter of controversial discussion [14–17].
So far, only t hree Fe-only nitrogenases h ave been
genetically (as anf s ystems) as well as biochemically
identified and characterized. These are the enzymes of
Azotobacter vinelandii [5,6], Rhodospirillum rubrum [9] and
Rhodobacter capsulatus [8,18,19], the heterometal-free
N
2
-fixation s ystem f rom the latter organism being t he most
intensively studied.
During the early years of F e nitrogenase r esearch, doubts
were widespread as to whether an Fe-only nitrogenase can
be isolated as an intact, functioning enzyme. These doubts
primarily arose due to the fact t hat preparations of the type
of anf-dependent nitrogenase were, regardless of their origin,
generally characterized by either extremely l ow catalytic
activity [5,6,9,18] or the wrong cofactor (namely the Fe Mo
cofactor) i ncorporated into the alternative d initrogenase
component [6,19]. H owever, a compr ehensive c haracteriza-
tion of the F e-only nitrogenase isolated from R. capsulatus,

in addition to H
+
), t he H
2
production rates w ere d istinctly
higher than the respective activities of the Mo nitrogenase
( sixfold). Samples of such highly active FeFe protein
preparations contained 26 ± 4 Fe atoms per protein
molecule, but neither m olybdenum nor vanadium [8].
A recent
57
Fe-Mo
¨
ssbauer-/Fe-EXAFS study on the FeFe
protein from R. capsulatus provided strong evidence that:
(a) the FeFe cofactor is diamagnetic in the Na
2
S
2
O
4
-
reduced state containing 4Fe
II
and 4Fe
III
centers, and (b) the
main structural feature of the FeMoco, the central trigonal
prismatic arrangement of Fe atoms, is also present in the
FeFe cofactor, thus indicating a s tructural homology

either the cofactor or the P-cluster has proven elusive due to
the fact that both of these metal clusters present in the
Fe-only n itrogenase are d iamagnetic in the dithionite-
reduced state, but probably become EPR-active upon
oxidation.
In the present work we focused on the identification or
tentative assignment of the most significant EPR signals
detected with FeFe protein samples, by pursuing the
following approach: t he FeFe and the MoFe proteins were
isolated from the s ame organism, samples were prepared
according to the same procedures and subsequently char-
acterized and compared b y E PR spectroscop y, particularly
with respect to their r edox properties.
MATERIALS AND METHODS
Bacterial strains
The o rganisms used were the R. capsulatus wild-type s train
B10S and the Mo-resistant double mutant with a nifHDK
deletion as well as an additional d eletion in t he modABCD
region [19,22]. The products of the l atter genes are involved
in high-affinity molybdenum transport [22].
Growth medium and culture conditions
The growth m edium and culture c onditions applied w ere as
described previously [8].
Purification of nitrogenase proteins
Preparation of cell-free extracts (cell disruption by lysozyme
followed by ultracentrifugation) were performed as des-
cribed by Sch neider et al.[8].Inviewofthedifficultyin
separating the dinitrogenase (Rc1
Mo
) and dinitrogenase

M
NaCl, w h ereas Rc2
Mo
was recovered in the 350 m
M
NaCl fraction. In the case of
theFenitrogenasetheRc2
Fe
component was eluted with
280 m
M
NaCl prior to the recovery of Rc1
Fe
with 330 m
M
NaCl. All nitrogenase component protein s were concentra-
ted t o approximately 8 mL by anaerobic ultrafiltration in a
50-mL chamber equipped with a PM30 Amicon membrane,
and subsequently fu rther c oncentrated to a final volume of
 1 mL in a B15 Amicon chamber. B oth dinitrogenase
components, which were of relevance for the present
comparative EPR study (Rc1
Mo
,Rc1
Fe
), were, based on
SDS/PAGE analysis, 90–95% pure.
The protocol previously employed to purify t he MoFe
protein (DEAE chromatography, Sephadex G-150 gel-
filtration) [8] led to a homogeneous preparation w ith

M
,pH7.4)at25°C in the presence of the f ollowing
mediators (each at 43 l
M
): 2,6-dichlorophenolindophenol,
phenazine methosulfate, thionine, methylene blue, indigo
trisulfonate, indigo carmine, resorufin, anthraquinone-
2-sulfonate, safranin O, benzyl viologen, methyl viologen.
Prior to the redox titration, the protein sample was
subjected to buffer exchange by gel filtration on Sephadex
G25 e quilibrated w ith 50 m
M
Hepes (pH 7.4) containing
1m
M
Na
2
S
2
O
4
(sodium dithionite). It is pertinent to note
that the reducing agent was not entirely removed from FeFe
protein preparations in view of the lability of the protein
even in the presence of only trace amounts of oxygen [8].
For the sake of direct comparison, MoFe protein samples
were treated under analogous con ditions.
The final sample solution (3 mL) containing 12–14 mg of
protein per mL was adjusted to different redox potentials by
the stepwise addition (0.5 lL) of K

M
NaClO
4
as an
external standard for integration.
RESULTS
EPR signals from oxidized states of the MoFe protein
In recent years EPR spectroscopic properties have been
reported for several MoFe proteins, mainly focusing on
P-cluster-type signals [27–31]. Based on the notion,
however, that, dependent on the origin, the purification
procedure and the s ample quality ( specific activity),
considerable differences within one class o f enzyme m ay
occur, we did not rely on literature data, but attempted
the direct experimental comparison of the MoFe and the
FeFe protein. We therefore isolated and prepared b oth
proteins not only from the same organism (R. capsulatus)
but also under the same conditions (lysozymatic cell
disruption, Q-Sepharose chromatography, E PR sample
preparation). For EPR experiments, protein samples were
used which displayed approximately m aximal specific
activities, i.e.  250 U (nmol a cetylene reducedÆmin
)1
)
per mg of FeFe protein and 1000–1200 UÆmg
)1
of MoFe
protein (compare [8]).
In the c ourse of thes e s tudies two e xperimental r outes to
obtain d ifferent redox states of the dinitrogenase protein

dithionite,
were supplemented with successively increasing amounts of
K
3
[Fe(CN)
6
], a rhombic S ¼ 1/2 EPR signal at g ¼ 2.07,
1.96 and 1.83 appeared (Fig. 1, spectrum 1). This signal was
1652 S. Siemann et al. (Eur. J. Biochem. 269) Ó FEBS 2002
most prominent w ith 2 m
M
K
3
[Fe(CN)
6
] and d ecreased
again above this concentration. With respect to its shape
and position of the g values, t his signa l appears t o
correspond to the S ¼ 1/2 signal that has been reported
for the partially oxidized MoFe proteins from Klebsiella
pneumoniae (Kp1) and A. vinelandii (Av1
Mo
) [28,30,32].
This type of signal has been interpreted to arise from the 1e
–
oxidized P-cluster (P
1+
)[28].
After the occurrence of an almost EPR-silent intermedi-
ate redox state (spectrum not shown), a second rhombic, but

organisms (e.g [27]), a midpoint potential (E
m
)of)50 mV
was determined for the S ¼ 3/2 FeMoco signal of Rc1
Mo
(Fig. 2). Above +100 mV the FeMoco signal disappeared
completely.
The EPR signal originating from the 1e
–
oxidized
P-cluster with the central g value a t 1.96 (in the following
designated as ÔP
1+
signalÕ) appeared at  )250 mV, reached
an intensity maximum at )120 mV and decreased a gain
with increasing potentials. The bell-shaped redox curve of
the P
1+
signal thus confirms the involvement of the
P-cluster in the transfer of at least two electrons (compare
[27,30]). The midpoint po tentials determined were:
)195 mV (E
m
for appearance of the signal representing
the P
N/1+
transition) and )30 mV (E
m
for disappearance;
P

O
4
. Spectrum 1, MoFe protein, oxidation with 2 m
M
K
3
[Fe(CN)
6
], measured at 16 K; spectrum 2, MoFe protein, oxidation
with 4 m
M
K
3
[Fe(CN)
6
], measured at 16 K; spectrum 3, FeFe protein,
oxidation with 2.5 m
M
K
3
[Fe(CN)
6
], measured at 10 K ; spectrum 4,
FeFe protein, oxidation with 2.5 m
M
K
3
[Fe(CN)
6
], measured at 23 K.

The axial signal and the rhombic P
1+
signal differed
significantly with respect to temperature and microwave -
power dependency. The P
1+
signal was most pronounced
around 18 K, the axial signal around 13 K. While the P
1+
signal appeared to be slightly power saturated already above
25 mW, the axial signal remained unsaturated even at
200 m W. However, both signals behaved similarly w ith
respect to their dependence o n the re dox potential. This
observation indicates that the axial signal might arise from
the P
1+
cluster as well, possibly i n a slightly modified
environment (protein conformation). I t is p ertinent to note
that this axial signal is a lso detectable in the spectrum
obtained a fter partial oxidation with K
3
[Fe(CN)
6
] withou t
mediators (at pH 7.4), although with much lower intensity
(data not shown).
The rhombic s ignal a t g ¼ 2.03, 2.00 and 1.90, which
appeared p rominently after oxidation w ith K
3
[Fe(CN)

Mo
this low fi eld sign al has been attributed to the
2e
–
-oxidized P-cluster (S ¼ 3) [27,30]. An exact determin-
ation of the midpoint potential was, however, not possible
due to the low intensity of this signal ( integer spin system)
under standard EPR conditions (perpendicular m ode).
The two characteristic
S
¼ 1/2 signals of the partially
oxidized FeFe protein
Stepwise oxidation of the FeFe protein. The p rotein
preparations used in this study contained 29 (± 3) Fe and
31 (± 4) acid-labile sulfur atoms. The high Fe/S content
indicates that these FeFe protein (Rc1
Fe
) preparations were
virtually devoid of any significant amounts of inactive
(oxidatively damaged clusters) or incompletely assembled
(vacant cofactor sites) enzyme. It is interesting t o note that
in the case o f dithionite-reduced VFe proteins [3,33] and
also in some instances with MoFe proteins [27,34] both such
protein forms gave ris e to S ¼ 1/2 s ignals. In sharp
contrast, the Rc1
Fe
protein is, in agreement w ith the
preceding report [8], apparently EPR silent in the presence
of excess dithionite. N either an S ¼ 3/2 nor a significant
S ¼ 1/2 signal in the g ¼ 2 region (< 0.05 spins/Rc1

protein, several novel EPR signals were detected. The two
most prominent signals (both S ¼ 1/2) have already been
partially c haracterized [8]. One of these is a very narrow
rhombic signal at g ¼ 2.00, 1.98 an d 1.96 ( in the f ollowing
designated as g ¼ 1.98 signal) and the other, a characteristic
broad signal with an absorption-shaped peak at g ¼ 2.27
and a derivative-shaped feature at g ¼ 2.06 (in the following
termed g ¼ 2.27 signal). The two signals are depict ed in
spectra 3 and 4 of Fig. 1 and directly c ompared to the most
characteristic S ¼ 1/2 signals of the reference system (the
oxidized MoFe protein), that h ave been attributed to P
1+
Fig. 3. pH-dependent occurrence of the axial EPR signal (g
||
¼ 2.00,
g
^
¼ 1.90) resulting from the partially oxidized MoFe protein. Two
samples of t he redox titration, b o th of t he potential region where the
rhombic P
1+
signal shows maximal intensity ()120 to )90 m V), were
thawed an d a djusted t o p H 6.4 and 8.4, r espe ctively, with a concen-
trated thre e-compon ent buffer system (0.8 7
M
Bistris, 0.44
M
Hepps,
0.44
M

were found to be catalytically intact (no loss of activity upon
oxidation). These results provide conclusive evidence that
the g ¼ 2.27 signal is not an art ifact. A s r egards t he n ature
of the signal, the lack of a visible negative a bsorption-
shaped peak at higher magnetic fields appears to be, at
first glance, indicative of an axial signal. However, a closer
inspection of the spectrum 3 in Fig. 4 reveals that the
derivative-shaped resonan ce at g ¼ 2.06 has approximately
the s ame intensity above and below t he baseline, suggesting
that the g ¼ 2.27 signal is rhombic. The inability to observe
the negative absorption-shaped peak may be the conse-
quence of inhomogeneous line broadening (g strain), a
phenomenon frequently observed in EPR spectra of met-
alloproteins [36].
InthecaseofAv1
Mo
, the typical P
1+
-cluster signal
was only detectable at neutral and weakly acidic pH, but
was absent at p H values near 8 .0 [30]. Because Rc1
Fe
samples were routinely prepared at pH 7.8, it was of
considerable interest to determine t he EPR properties
also under weakly acidic conditions. In fact, EPR spectra
of thionine-oxidized samples prepared at pH 8.4, 7.4 and
6.4 (Fig. 4) revealed a n ew signal, which was most
pronounced at pH 6.4, but proved to be very similar to
the g ¼ 2.27 feature. The signal was slightly less broad
corresponding to a shift of the g values to 2 .22 and 2.05

ences in the cluster environment or to protonation/depro-
tonation effects in the cluster itself. pH-based signal shifts
and the occurrence of a dditional signals have been reported
for the cofactor of the MoFe proteins from K. pneumoniae
(Kp1) [38], Xanthobacter autotrophicus (Xa1) [39] and even
for the isolated FeMoco from Av1
Mo
[40].
The strong resonance near g ¼ 2.00, present in all spectra
of Fig. 4, originates from the t hionine radical s ignal. Under
the oxidation conditions (10 m
M
thionine) applied in these
experiments, the narrow g ¼ 1.98 signal was absent or of
such low intensity that it was completely obscured by the
radical signal. Analogous experimen ts on pH-dependence
with samples oxidized w ith K
3
[Fe(CN)
6
]revealedthat
the n arrow signal w as not significantly influenced by the
pHvalue (data not shown). For the clarity of presentation
of the broad-type signals, we chose the spectra of the
Fig. 4. pH-dependent shift of the broad g ¼ 2.27 EPR signal under
slightly acidic conditions. A FeFe protein sample (9 mgÆmL
)1
), freshly
prepared in the presence of 4 m
M

discussed in a later section. In addition, the quantification of
the presumably rhombic g ¼ 2.27 signal is solely based on
the g ¼ 2.27, 2.06 peaks and did not include the putative
thirdpeakintheg ¼ 1.8 region, thus leading to an
underestimation of t he spin con tent. The rhombic signal at
g ¼ 2.22, measured at pH 6.4, was not quantified because
of its generally low er intensity as well as the decreased
stability of the protein at this pH value. The narrow
g ¼ 1.98 sign al integrated to only 0.25 spins per protein
molecule, indicating that the c orresponding cluster, at least
in this specific redox state, is not of catalytic relevance.
Equilibrium-mediated redoxtitration of the FeFe pro-
tein. A redox titration (at pH 7.4) of the broad g ¼ 2.27
signal in the p resence of m ediators resulted in a be ll-shaped
titration curve with midpoint potentials of )80 mV
(appearance) and +70 mV (disappearance). Maximal
signal intensity was achieved by adjusting the potential to
)5 mV (Fig. 5). These results imply that the cluster giving
rise to the g ¼ 2.27 signal is, in analogy to the P
1+
cluster of
the MoFe protein, involved in a 2e
–
transfer process. It
could be reversibly converted into an EPR-silent state either
by reduction or by further o xidation.
The narrow g ¼ 1.98 signal was observed in a region
shifted about 150 mV to more positive redox potentials.
For t his signal a midpoint potential of 80 mV (appear ance)
was determined (Fig. 5). Maximal signal intensity was

Fe
,theg ¼ 4.3 s ignal significantly
increased towards the e nd of the redox titration
(+220 mV).
(b) Preliminary studies on the isolation and purification
of the FeFe apoprotein (from a nifBB’strain) revealed
that the cofactorless protein, when prepared according to
the procedure approved for the native enzyme [8], cannot
be obtained in an intact hexameric, but only in a
tetrameric a
2
b
2
form. The small d subunit could b e
isolated by DEAE chromatography as a separate peptide
(D. Tiemann, S. Fuchs, K. Schneider & A . Mu
¨
ller,
unpublished results). Dissociation of the d subunit from
the apodinitrogenase under certain conditions (e.g. during
gel filtration) has also been reported in the case of the
vanadium nitrogenase (VFe protein) from A. vinelandii
[42]. The tetrameric FeFe apoprotein from R. capsulatus
did not show any EPR signal typical of a P-cluster signal.
Only a signal at g ¼ 2.01, rather reminiscent of an
[Fe
3
S
4
]

performed w ith the resonance at g ¼ 1.9 6. Spectra were m easured at
23 K a nd 100 mW.
1656 S. Siemann et al. (Eur. J. Biochem. 269) Ó FEBS 2002
Comparison of the turnover signals from enzyme-
reduced states of the MoFe protein and the FeFe protein
The characteristic EPR signal of the FeFe protein (Rc1
Fe
)
under turnover conditions at g ¼ 1.96, 1.92 and 1.77 has
already been documented [8]. The occurrence of this
S ¼ 1/2 t ype signal with samples containing ATP, a
substrate (N
2
,C
2
H
2
,H
+
) and the dinitrogenase reductase
component for enzymatic reduction of the FeFe protein,
was confirmed in the present study (Fig. 6A, spectrum 1).
The signal was most prominent in the presence of the
natural substrate, N
2
,whenmeasuredat16Kand100mW.
In order to avoid (a) P-cluster oxidation ( see D iscussion
section on this t opic) and (b) an interference of t he turnover
signal and t he dinitrogenase r eductase (Rc2
Fe

,Rc2
Mo
)
according to the same procedures as the Fe nitrogenase
components (Rc1
Fe
,Rc2
Fe
) and finally applied exactly
identical turnover and EPR conditions, t he EPR s ignal
detected with the Fe-only nitrogenase ( g ¼ 1.96, 1.92,
1.77) at 16 K, was absent (Fig. 6A, spectrum 2). The
minimal resonance (g ¼ 1.9–2.0) visible in s pectrum 2
resulted from Rc2
Mo
. It w as identical with the control
spectrum in the absence of Rc1
Mo
.Furthermore,withthe
exception of the classical S ¼ 3/2 signal of the MoFe
protein a t l ower temperatures (< 12 K), no other signal
was detectable. In full accordance with literature data ( e.g
[44]), the e nzyme(Rc2
Mo
)-reduced state o f t he MoFe
protein showed, compared t o the dithionite-reduced state
(Fig. 6B, spectrum 3), a drastic decrease ( 70%) in signal
intensity ( Fig. 6B, spectrum 4). In th e case of Mo n itro-
genases f rom o ther organisms, t his b ehaviour has been
interpreted to be due to one-electron reduction of the

Fig. 6. EPR signals o f the FeFe protein and the MoFe protein under
turnover conditions. (A) EPR sp ectra of b oth the FeFe protein ( spec-
trum 1) and the MoFe prote in (spectrum 2) me asured at 16 K under
turnover conditions. The sam ples were prepared anaerobically (under
N
2
) directly in the EPR tube. They contained 2 4 m g Rc1
Fe
per mL and
0.6 mg Rc2
Fe
per mL in the case of the F e nitrogenase and 28 mg
Rc1
Mo
per mL and 0.7 mg Rc2
Mo
per mL in the case of the Mo
nitrogenase. The other constituents were: 100 m
M
Hepes (pH 7.8),
5m
M
ATP, 10 m
M
MgCl
2
,6 m
M
Na
2

mL but no Rc2
Mo
. All other conditions were equal to those described
for the tu rn over samples. Both spectra w ere recorded at 2 0 mW.
Ó FEBS 2002 Redox properties of the FeFe protein (Eur. J. Biochem. 269) 1657
S ¼ 3/2 FeMoco signal (E
m
 )50 mV), of a characteristic
rhombic S ¼ 1/2 EPR signal at g ¼ 2.07, 1.96, 1.83, appar-
ently deriving from the one-electron-oxidized P-cluste r (P
1+
)
and of a weak signal near g ¼ 12 (propably S ¼ 3),
reminiscentofthe2e
–
oxidized P-cluster (P
2+
) signal [27,30].
However, we have found that Rc1
Mo
significantly
deviates from other MoFe proteins with respect to a
number of relevant characteristics.
(1) Although the lineshape and g value positions (2.07,
1.96, 1.83) of t he EPR s ignals from Rc1
Mo
and Av 1
Mo
,
which have been interpreted to represent the P

the analogous state of the Rc1
Mo
P-cluster.
(c) The intensity of the rhombic S ¼ 1/2 P
1+
signal of
Av1
Mo
is strongly pH-dependent as well, being maximal a t
pH 6.0. At pH 7.5 this signal is of only very low intensity
and at pH 8.0 it is even absent. This observation might
explain why the characteristic P
1+
cluster signal has not
been detected by some research groups [27]. The absence of
the P
1+
state in a weakly alkaline medium is likely to be
caused by the simultaneous transfer of two electrons, thereby
resulting in a transition from P
N
directly into the P
2+
state.
In the case of the rhombic P
1+
signal of Rc1
Mo
(g ¼ 2.07,
1.96, 1.83), which corresponds to the Av1

–
oxidized P-cluster (P
3+
) [27]. A
rhombic signal in the same potential range has also been
observed in t he case of Rc1
Mo
, however, t his signal was less
broad a nd located in a lower m agnetic fi eld region
(g ¼ 2.03, 2.00, 1.90). Moreover, the R c1
Mo
signal showed
significant intensity only after nonmediated oxidation and
did not, in contrast to the Av1
Mo
signal, disappear upon
further oxidation with K
3
[Fe(CN)
6
](>300mV).
(4) The phenomenon of Ôspin mixtures Õ, i.e. the simulta-
neous occurrence of S ¼ 1/2, 5/2 signals (P
1+
state) [28] and
of S ¼ 1/2, 7/2 signals (P
3+
state) [27], was not observed
with the Rhodobacter enzyme. T he observation that P-
clusters of one and the same redox state m ay be present in

evidence that this signal is not an artifact or caused by a
Rhodobacter-specific paramagnetic impurity.
(b) Spectroscopic studies (Mo
¨
ssbauer, Integer-spin-
EPR) on MoFe proteins h ave revealed that all iron
atoms in the dithionite-reduced P-cluster are most likely
in the ferrous state [47,48]. This excludes the possibility
that the diamagne tic, fully reduced P -cluster becomes
further reduced during enzyme turnover. Thus, the
turnover signal of the FeFe protein cannot arise from
a ‘super-reduced’ P-cluster.
Table 1. A comparative overview on the EPR s ignals of the M oFe- and FeFe pro tein from R. capsulatus. The enzyme (dinitrogenase r eductase )-
reduced MoFe protein mo lecu les (E
1
state) are EPR-silent. However, under the con ditions used, i.e. at low electron flux (Rc2
Fe
:Rc1
Fe
¼ 1 : 10), the
sample also contained  30% o f dithionite-reduced M oFe protein molecules (Ôresting stateÕ E
0
) showing the typical S ¼ 3/2 FeMoco s ignal.
Redox state
EPR signals (g values)
FeFe protein MoFe protein
I. Enzyme-reduced 1.96, 1.92, 1.77 EPR-silent
II. Na
2
S

3
and
E
4
,postulatedtoplayaroleinN
2
-binding/reduction and
suggested to be connected to P-cluster oxidation, are
presumably present only in very small proportions, unde-
tectable by EPR under these conditions. These theoretical
expectations are in full accordance with the results
obtained with the reference system used in this study, the
Mo nitrogenase f rom R. capsulatus. Upon enzymatic
reduction of the MoFe protein at a R c2
Mo
/Rc1
Mo
ratio
of 1 : 10 the FeMoco EPR s ignal (representing the E
0
state) maintained 30% of its intensity. In addition, no
signal indicative of an oxidized P-cluster was detected.
Under analogous conditions, the Fe-only nitrogenase failed
to exhibit a ny of the signals appearing upon oxidative
titration. Consequently, the turnover signal does not arise
from an oxidized P-cluster.
(d) As concluded from the results of the preceding
Mo
¨
ssbauer study [21], the FeFe cofactor of the Fe-only

principle the same (1.90–2.03).
(b) The redox potentials at which the narrow FeFe
protein signal occurs (E
m
¼ 80 mV) and reaches maximal
intensity ( 160 mV), are in excellent agreement with the
values reported for the P
3+
-cluster signal of the MoFe
protein from A. vinelandii [27].
(c) After being induced by oxidation w ith ferricyanide,
both clusters giving rise to this type of signal can only
partially ( 20–30%) b e re-reduced by the addition of
excessive amounts of dithionite.
Although a three-electron-oxidized P-cluster can appar-
ently be produced by chemical oxidation, the irreversibility
of this in vitro process a s well as the low spin content of the
corresponding EPR signal i ndicate that the P
3+
state i s not
of physiological/catalytical relevance.
The assignment of the novel broad g ¼ 2.27 feature of
the partially oxidized FeFe protein appears to be even
more challenging. In fact, such a signal has never been
observed for any type of FeS cluster. The characteristic
g values of all known S ¼ 1/2 systems arising from
homonuclear FeS clusters are situated between 1.8 and
2.15 [51].
Several fundamental considerations oppose the attribu-
tion of this signal to an oxidized P-cluster:

(b) The possibility that the g ¼ 2.27 signal represents P
2+
appears t o be highly unlikely a s well. The 2e
–
-oxidized
P-cluster of M oFe p roteins h as been reported not to reveal
an S ¼ 1/2 signal, but to exhibit a signal at g  12, resulting
from an integer spin state (presumably S ¼ 3[27]).
Although this feature has b een demonstrated in the c ase
of Rc1
Mo
as well, a corresponding signal for the Rc1
Fe
protein was not detected. Under the EPR spectroscopic
conditions employed in this study (perpendicular mode), the
2e
–
oxidized P-cluster of R c1
Fe
is EPR-silent.
(c) If our conclusion is correct that the narrow g ¼ 1.9 8
signal arises indeed from the 3e
–
oxidized P-cluster, the
possibility that the broad g ¼ 2.27 signal represents P
1+
(or
P
2+
) can automatically be excluded in v iew of the following

Ó FEBS 2002 Redox properties of the FeFe protein (Eur. J. Biochem. 269) 1659
the high potential region at which this narrow signal occurs,
this possibility appears very unrealistic.
In summa ry, if t he g ¼ 2.27 signal does not arise from
P
1+
or P
2+
, th e only alternative is that it results from the
FeFe cofactor. Because evidence has been provided by EPR
and Mo
¨
ssbauer spectroscopy that the FeFeco is diamag-
netic in the dithionite-reduced state, it appears plausible that
the g ¼ 2.27 signal might represent the one-electron-oxid-
ized state of t he cofactor. T he observation of a bell-shaped
redox curve (Fig. 4) implies that the cluster giving rise to this
signal can u ndergo further oxidation. This seems t o be
somewhat surprising, since the EPR-silent 1e
–
-oxidized
FeMoco cannot be further oxidized to another E PR
detectable state. Although such a redox state appears
unlikely to participate in th e catalytic events leading to
substrate r eduction, its f ormation in the case of the Fe-only
nitrogenase might be explained by the involvement of the
additional iron a tom (which replaces the heterometal atom)
in that redox process.
As regards the lineshape and width of the g ¼ 2.27
signal, its unusual broadness may arise from Heisenberg

1+
transition. As already outlined above,
such a 2e
–
transfer, responsible for the absence of a
characteristic P
1+
EPR-signal, has been postulated to occur
inthecaseoftheAv1
Mo
protein as well, however, only at
weakly alkaline pH [27,30].
In conclusion, the assignment of the characteristic EPR
signals of the FeFe protein is tentative. Future studies will
therefore aim at a c onclusive identification of t he EPR
signals, particularly of those arising from the oxidized states
of the P-cluster and the FeFeco. Such investigations will
include the development of a procedure for the isolation and
stabilization of a hexameric FeFe apoprotein, as well as a
detailed EPR spectroscopic characterization of this cofac-
torless protein system.
ACKNOWLEDGEMENTS
The authors are indebted to Mrs S . Selsemeier-Voigt for technical
assistance. This work was supported by the Deutsche Forschungs-
gemeinschaft (DFG).
REFERENCES
1. Hales, B.J., Case, E.E., M orningstar, J.E., Dzeda, M.F. &
Mauterer, L.A. (1986) Isolation of a new vanadium-containing
nitrogenase f rom Azotobacter vinelandii. Biochemistry 25, 7251–
7255.

Rhodobacter c apsulatus. Eur. J. Biochem. 24 4, 789–800.
9. Davis,R.,Lehmann,L.,Petrovich,R.,Shah,V.K.,Roberts,G.P.
& Ludden, P.W. (1996) Purification and characterization of the
alternative nitrogenase from the p hotosynthetic bacterium.
Rhodospirillum rubrum. J. Bacteriol. 17 8, 1445–1450.
10. Ribbe, M., Gadkari, D. & Meyer, O. (1997) N
2
fixation by
Streptomyces thermoautotrophicus involves a molybdenum-dini-
trogenase and a manganese-superoxide oxidoreductase that cou-
ples N
2
reduction to the o xidation of s uperoxide produced from
O
2
by a molybdenum-CO dehydrogenase. J. Biol. Chem. 272,
26627–26633.
11. Hawks, T.R., McLean, P.A. & Smith, B .E. (1984) Nitrogenase
from nifV mu tants of Klebsiella pneumoniae contains an altered
form of the iron-molybden um cofactor. Biochem. J. 217, 317–321.
12. Peters, J.W., Stowell, M.H.B., Soltis, S.M., Finnegan, M.G.,
Johnson, M.K . & R ees, D.C. (1997) Redox-dependent structural
changes in the nitrogenase P-cluster. Biochemistry 36, 1181–1187.
13. Howard, J.B. & Rees, D.C. (1994) Nitrogenase: a nucleotide-
dependent molecular switch. Annu. Rev. Biochem. 63 , 235–264.
14. Deng, H . & Hoffmann, R. (1993) How N
2
might be a ctivated by
the FeMo-cofactor in nitrogenase. Angew. Chem. Int. E d. 32,
1026–1029.

ddekopf, K. & Klipp,
W. (1993) Detection of the in vivo incorporation of a metal cluster
into a protein. The FeMo cofactor is inserted into the FeFe protein
1660 S. Siemann et al. (Eur. J. Biochem. 269) Ó FEBS 2002
of the alternative nitrogenase of Rhodobacter capsulatus. Eur.
J. Biochem. 215, 25–3 5.
20. Krahn,E.,Weiss,B.J.R.,Kro
¨
ckel, M ., Cramer, S .P., Trautwein,
A.X., Schneider, K. & Mu
¨
ller, A. (1998) The Fe-only nitrogenase
from Rhodobacter capsulatus: 2. The F eFe p rotein metal c enters
probed by EXAFS and Mo
¨
ssbauer spectroscopy. In Biological
Nitrogen Fixation for the 21st Century (Elmerich, C., K ondo rosi,
A. & Newton, W.E., e ds), pp. 59–60. Kluwer Academic Publish-
ers, Dordrecht, the Netherlands.
21. Krahn, E., Weiss, B.J.R., Kro
¨
ckel, M., Groppe, J., Henkel, G.,
Cramer,S.P.,Trautwein,A.X.,Schneider,K.&Mu
¨
ller, A. (2002)
The F e-only ni trogenase f rom Rhodobacter capsulatus: identifica-
tion of the cofactor, an unusual, high-nuclearity iron-sulfur
cluster, by Fe K-edge EXAFS and
57
Fe Mo

sulfur protein sampl es containing dithionite. Ana l. Biochem. 79,
157–165.
26. Dutton, P.L. (1978) Redox potentiom etry: determination of
midpoint potentials of oxidation-reduction components of bio-
logical electron-transfer sy stems. Methods E nzymol. 54, 411–435.
27. Pierik, A.J., W assink, H., Haaker, H . & Hagen, W.R. (1993)
Redox properties and EPR spectroscopy of the P clusters of
Azotobacter vinelandii MoFe protein. Eur. J. Biochem. 212, 51–61.
28. Tittsworth, R.C. & Hales, B.J. (1993) Detection of EPR signals
assigned to the 1-equivalent-oxidized P-clusters of the nitrogenase
MoFe protein from Azotobacter vinelandii. J. Am. Chem. Soc. 115,
9763–9767.
29. Tittsworth, R.C. & Hales, B.J. (1996) Oxidative titration of the
nitrogenase VFe p rotein from Azozobacter vinelandii: an example
of redox-gated electron flow. Bioc hemistry 35, 4 79–487.
30. Lanzilotta, W.N., Christiansen, J., Dean, D.R. & Seefeldt, L.C.
(1998) Evidence for coupled electron and proton transfer in the
[8Fe)7S] cluster of nitrogenase. Bioc hemistry 37, 1137 6–11384.
31. Chan, J.M., Dean, D.R. & Seefeldt, L.C. (1999) Spectroscopic
evidence for changes in the redox state of the nitrogenase P cluster
during turnover. Biochemistry 38, 5779–5785.
32. Smith, B.E., Lowe, D.J., Chen, G X., O’Donell, M.J. & Hawkes,
T.R. (1983) Evidence on intramolecular electron transfer in the
MoFe protein of nitrogenase from Klebsiella pneumoniae from
rapid-freeze electron-paramagnic-resonance studies of its oxida-
tion by ferricyanide. Biochem. J. 209, 207–213.
33. Blanchard, C.Z. & Hales, B.J. (1996) Isolation of two forms of the
nitrogenase VFe protein from Azotobacter vinelandii. Biochemistry
35, 472 –478.
34. Zumft, W.G. & Mortenson, L.E. (1973) E vidence for a catalytic-

40. Newton,W.E.,Gheller,S.F.,Feldman,B.J.,Dunham,W.R.&
Schultz, F.A. (1989) Isolated iron-molybdenum cofactor of
nitrogenase exists in multiple forms in its oxidized a nd semi-
reduced states. J. Biol. C hem. 264, 1924 –1927.
41. Palmer, G. (1985) The electron paramagnetic resonance of
metalloproteins. Bioc hem. Soc. Trans . 13, 548 –560.
42. Chatterjee, R ., Ludden, P.W. & S hah, V.K. (1 997) Characteriza-
tion of VNFG, the d subunit of the vnf-encoded apodinitrogenase
from Azotobacter vinelandii. Implications for its role in the forma-
tion of functional dinitrogenase 2. J. Biol. Chem. 272, 3758–3765.
43. Zaborosch, C., Ko
¨
ster, M., Bill, E., Schneider, K. & Schlegel,
H.G.&Trautwein,A.X.(1995)EPRandMo
¨
ssbauer spectro-
scopic stud ie s on the tet rameric, NAD-linked h ydrogenase of
Nocardia opaca 1b and its two dimers: 1. The bd-dimer – a pro-
totype of a simple h ydrogenase. Biometals 8, 149–162.
44. Smith,B.E.,Lowe,D.J.&Bray,R.C.(1973)Studiesbyelectron
paramagnetic resonance on the catalytic mechanism of
nitrogenase o f Klebsiella pneumoniae. Biochem . J. 135, 331–341.
45. Burgess, B.K. (1990) The iron–molybdenum cofactor of nitro-
genase. Chem. Rev. 90 , 1377–1406.
46. Onate, Y.A., Finnegan, M.G., Hales, B.J. & Johnson, M.K.
(1993) Variable temperature magnetic circular dichroism studies
of reduced nitrogen ase iron proteins an d [4Fe)4S]
+
synthetic
analog clusters. Bi ochim. Biophys. Acta 116 4, 113–123.

Ó FEBS 2002 Redox properties of the FeFe protein (Eur. J. Biochem. 269) 1661