© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
ARTICLES
The structure of bovine F1-ATPase in complex with its
regulatory protein IF1
Elena Cabezón1,3, Martin G Montgomery1, Andrew G W Leslie2 & John E Walker1
In mitochondria, the hydrolytic activity of ATP synthase is prevented by an inhibitor protein, IF1. The active bovine protein
(84 amino acids) is an -helical dimer with monomers associated via an antiparallel -helical coiled coil composed of residues
49–81. The N-terminal inhibitory sequences in the active dimer bind to two F1-ATPases in the presence of ATP. In the crystal
structure of the F1–IF1 complex at 2.8 Å resolution, residues 1–37 of IF1 bind in the DP-DP interface of F1-ATPase, and also
contact the central subunit. The inhibitor opens the catalytic interface between the DP and DP subunits relative to previous
structures. The presence of ATP in the catalytic site of the DP subunit implies that the inhibited state represents a pre-hydrolysis
step on the catalytic pathway of the enzyme.
ATP synthase (also called F1Fo–ATPase) is a multisubunit, membranebound assembly central to biological energy conversion. It is composed
of a globular F1 catalytic domain (subunit composition α3β3γδε) and an
Fo domain, linked together by central and peripheral stalks1–3. Energy
from the transmembrane proton-motive force (PMF) generates the rotation of a ring of hydrophobic c subunits in the Fo domain against another
hydrophobic protein, subunit a (or ATPase–6). ATPase-6 is probably
linked via subunit b and the peripheral stalk (subunits b, d, F6 and
OSCP)4 to the external surface of the F1 domain5,6. The central stalk is
associated with the c ring via its foot7, and the rotation of the c ring makes
the central stalk turn with it. The central stalk consists of the γ, δ and
ε subunits8 and the γ subunit contains an asymmetric coiled coil of
α-helices, which penetrates through the central axis of the F1 domain8,9.
The three α and three β subunits are arranged in alternation around this
α-helical coiled coil, and the three catalytic sites of the enzyme lie in
β subunits at interfaces with α subunits. Because of the asymmetry of the
coiled coil, the three catalytic sites have different nucleotide binding
properties. In the original crystal structure of the bovine enzyme9, which
will be referred to as the reference state, one subunit (βDP) binds ADP
(corresponding most closely to the ‘tight’ state of the binding change
mechanism1), the second (βTP) binds ATP (and corresponds to the
‘loose’ state) and the third (βE) is in a conformation that has a very low
affinity for a nucleotide (corresponding to the ‘empty’ or ‘open’ state).
During ATP synthesis, a 120° rotation of the central stalk brings about an
interconversion of sites such that the ‘open’ site becomes ‘tight’, the ‘loose’
site becomes ‘open’ and the ‘tight’ site becomes ‘loose’. Thus, each 360°
rotation of the central stalk, in 120° steps, takes each catalytic site through
the three states, and three ATP molecules are synthesized.
When a cell is deprived of oxygen, for example by ischemia, the PMF
across the inner mitochondrial membrane collapses and the ATP
synthase switches from synthesis to hydrolysis. This hydrolytic activity is
prevented by a natural inhibitor protein, IF1 (refs. 10,11). Its action
depends on the presence of ATP, and in the inhibited complex an ATP
molecule becomes trapped in the enzyme. Residues 14–47 of the 84residue bovine IF1 have been defined by deletion analysis as the minimal
inhibitory sequence12. This region is 87% identical in the bovine and
human proteins. In vitro, the active form of bovine IF1 is a dimer associated by formation of an antiparallel α-helical coiled coil between the
C-terminal regions of monomers13,14. The dimeric IF1 binds simultaneously to two F1 domains via its inhibitory regions15. The dimeric state
of the inhibitor is favored by pH values <6.5 (ref. 16). Under ischemic
conditions in vivo, glycolysis becomes the only source of cellular ATP. The
pH values of the cytosol and the mitochondrial matrix decrease17–19,
promoting inhibition of ATP hydrolysis by IF1 to preserve ATP. Upon
re-energization of mitochondria, the pH increases, dimeric IF1 dissociates
from F1 and, in vitro, it forms inactive tetramers and higher oligomers via
coiled-coil interactions in the N-terminal inhibitory regions14.
This paper describes the structure of the inhibited complex between
F1-ATPase and IF1 formed in the presence of ATP. The structure confirms
that the N-terminal region of the dimeric inhibitor is bound to F1. It
exerts its effect by binding selectively to the interface between the αDP and
βDP subunits of F1, thereby blocking rotary catalysis. In the inhibited
state, the enzyme is locked in a conformation in which ATP (or AMPPNP) is entrapped in subunit βDP. The structure of the complex seems to
represent a pre-hydrolysis state in the catalytic cycle.
RESULTS
Disorder in the crystals of the F1–IF1 complex
The structure of the bovine F1–IF1 complex was solved by molecular
replacement and refined against X-ray data to a resolution of 2.8 Å
1The
Medical Research Council Dunn Human Nutrition Unit, Hills Road, Cambridge CB2 2XY, UK. 2The Medical Research Council Laboratory of Molecular Biology,
Hills Road, Cambridge CB2 2QH, UK. 3Present address: Departamento De Biologia Molecular, Universidad De Cantabria, Santander, Spain. Correspondence should
be addressed to J.E.W. ([email protected]) or A.G.W.L ([email protected]).
Published online 17 August 2003; doi:10.1038/nsb966
744
VOLUME 10 NUMBER 9 SEPTEMBER 2003 NATURE STRUCTURAL BIOLOGY
ARTICLES
© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
a
b
In regions that deviate significantly from
three-fold pseudo symmetry, such as parts
of the nucleotide binding domain and the
C-terminal domain of the βE subunit,
the density is weaker, but still generally
unambiguous. The quality of the electron
density for the inhibitor and the βDP catalytic
site is apparent in Figure 2.
The disorder observed in the crystals results
from the dimerization of F1-ATPase by the
inhibitor protein15. The distance between two
F1-ATPase complexes related by a unit cell
translation along the crystallographic b-axis is
such that they can be cross-linked by a dimeric
inhibitor molecule, but only if the second
F1-ATPase complex is rotated by 120° relative
to the first. If cross-linked F1-ATPase dimers
were arranged perfectly in the crystal, this
arrangement would give rise to a doubling in
length of the b-axis. In reality, the packing of
F1-ATPase dimers in the lattice is imperfect,
resulting in a statistically disordered structure
that gives rise to the continuous diffraction
observed in the diffraction images.
Overall protein structure
The two F1-ATPase complexes adopt similar
but not identical conformations. The overall
r.m.s. deviation in Cα coordinates is 0.82 Å.
Significant differences are restricted to the αE
and αDP subunits, which show small rigidc
body rotations. Because these subunits are
involved in intermolecular contacts, it is likely
that the rotations are the result of crystal
packing forces.
The structure of dimeric F1–IF1 (Fig. 1a,b)
contains 5,980 amino acids. The region
defined as complex A consists of residues
αE24–510, αTP24–401 and αTP410–510,
αDP19–404 and αDP409–510, βE9–474,
βTP9–474, βDP9–477, γ1–30, γ77–88 and
Figure 1 The dimeric bovine F1-ATPase–IF1 complex. (a) View of the dimeric complex along the central
axes of F1 particles viewed from the protruding region of the central stalk toward the catalytic sites.
γ221–272 and residues 4–40 of IF1. Complex B
Complexes A and B (left and right, respectively) are shown in ribbon representation. The α, β and γ
includes the same residues as complex A,
subunits are red, yellow and blue, respectively, and residues 4–40 and 4–47 of the two IF1 molecules
except that residues αDP400–411 are absent
are green. The green arrows indicate that the N-terminal regions of IF1 are linked via an antiparallel
and residues 41–47 of IF1 were resolved. The
coiled coil in their C-terminal regions that is not resolved in the map. (b) Space-filling representation of
δ and ε subunits could not be modeled in
a side view using the same color scheme as in a. (c) Ribbon representation of the structure of isolated
either complex. The ordered region of the
IF1 dimers14 shown alone (identical monomers are red and blue) and with the superimposed N-terminal
segments (in green), representing residues 4–40 and 4–47 from complexes A and B, respectively.
inhibitor protein adopts a helix-turn-helix
structure, with at least residues 4–37 involved
directly in contact with F1. There was no inter(Table 1). The crystals of the complex are statistically disordered so pretable density for the C-terminal regions of IF1 (residues 48–84),
that, on average, each site in the crystal is occupied by two complexes which form an antiparallel coiled coil in isolated dimeric IF1. IF1 in the
of half occupancy. These two complexes, A and B, are in different ori- dimeric F1–IF1 complex (Fig. 1c, green) is superimposed on the strucentations and are related by a rotation of 120° around the axis of ture of dimeric IF1 alone14 (blue and red). The r.m.s. deviations in Cα
pseudo symmetry of the (αβ)3 assembly. This arrangement super- coordinates in the overlapping segments (residues 19–40 in complex A
imposes the βDP, αDP, βTP, αTP, βE and αE subunits of F1-ATPase and residues 20–47 in complex B) are 0.99 Å and 0.92 Å, respectively. In
complex A onto the βTP, αTP, βE, αE, βDP and αDP subunits, respect- the F1–IF1 complex, residues 4–18 of IF1 were resolved, whereas in the
ively, of complex B (Fig. 1). Because a significant proportion of structure of the IF1 dimer alone they were disordered. The distance
F1-ATPase follows three-fold pseudo symmetry, the resulting electron between the two inhibitory regions in the isolated IF1 dimer in the fully
density for the superimposed structures is of reasonable quality, much extended form is 62 Å (Fig. 1c), whereas in the crystal structure of the
like that obtained for a fully ordered structure at a resolution of ∼3.5 Å. F1–IF1 complex, it is reduced to 31 Å. Crystals of the F1–IF1 complex
NATURE STRUCTURAL BIOLOGY VOLUME 10 NUMBER 9 SEPTEMBER 2003
745
ARTICLES
© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
a
b
Figure 2 Regions of the electron density map of the F1–IF1 complex. (a) Averaged electron density map at 2.8-Å resolution calculated with CNS from
residues 18–34 of the IF1 protein. (b) Stereo view of the 2.8 Å resolution 2Fo – Fc (blue) and Fo – Fc (green) electron density maps calculated with CNS
(contoured at 1 σ and 2.2 σ, respectively) from the nucleotide binding site in the βDP subunit of F1-ATPase. To avoid model bias, this map was calculated
at a stage of refinement before inclusion of the γ-phosphate in the model.
were redissolved and examined by SDS-PAGE. The IF1 was intact, and
there was no evidence of its degradation in the crystals. Therefore, the
unresolved region must have a greater curvature than in the crystals of
the IF1 dimer and also be disordered. In the crystals of IF1 alone, two
types of dimers were observed with markedly different curvatures in
this region, suggesting that the coiled-coil region is flexible14.
Interactions between F1-ATPase and IF1
Bovine IF1 binds to F1 at the interface between the αDP and βDP subunits, and it also makes contacts with the γ subunit and with residues
βTP386 and αE355 (Fig. 3). The extensive binding surface has an area
of 1,500 Å2 and is a mixture of 60% hydrophobic and 40% hydrophilic
interactions. Residues αDP405–408 may also contribute to binding IF1,
but the electron density did not allow this region to be modeled.
Earlier, the interactions between yeast F1Fo–ATPase and its inhibitor
protein were shown by cross-linking and peptide mapping experiments to involve residues at an interface between α and β subunits20.
On the basis of similar experiments with the bovine F1-ATPase, it was
proposed that a region within residues β394–459 interacts with IF1
(ref. 21). Both observations are consistent with the X-ray structure.
By deletion analysis, residues 14–47 of bovine IF1 were defined as
the smallest segment to retain full inhibitory activity, except that,
unlike the situation with intact IF1, the inhibition was lost with time,
presumably because the peptide was dissociating from the enzyme12.
The structure shows that residues 1–13 make intimate contacts with
F1-ATPase and thus stabilize the binding of IF1 to F1. Residues 48–56 of
IF1 also helped to stabilize the F1–IF1 complex12. However, they do not
interact directly with F1-ATPase, and they probably contribute to
stability by formation of IF1 dimers in the coiled-coil region14.
Conformational changes in F1-ATPase
The reference state structure of F1-ATPase was determined with crystals grown in the presence of AMP-PNP and ADP9, whereas the present structure was determined with crystals of the F1–IF1 complex
preformed in the presence of ATP, and then grown in the presence of
AMP-PNP. In the structure of the complex, the catalytic sites in subunits βTP and βDP are occupied by ATP or AMP-PNP (which cannot
be distinguished on the basis of the electron density), whereas subunit
βE is unoccupied. Either ATP or AMP-PNP is bound to each of the
three α subunits. The presence of ATP (or AMP-PNP) rather than
ADP in the βDP subunit is very clear (Fig. 2b) and is presumably a consequence of IF1 binding, because AMP-PNP was present at high concentration (250 µM) in the crystallization of the reference state
structure. The conformational changes in F1 due to inhibitor binding
were analyzed by comparison of structures of the reference state of F1
with the F1–IF1 complex. For this purpose, a common point of reference was defined as the six N-terminal domains of the (αβ)3 subassembly, which form a stable crown to the structure. After
superimposing all six corresponding
N-terminal domains, changes in quaternary
and tertiary structure were demonstrated by
examining the r.m.s. deviations in Cα positions for each domain. As expected, the
N-terminal domains are very similar (r.m.s.
deviations up to 0.5 Å), and the largest r.m.s.
deviations are in the C-terminal domains of
the αDP and βDP subunits (4.5 Å and 3.0 Å,
respectively). They arise from rotations of
Figure 3 Stereo view of the contacts between
F1-ATPase and IF1. Side chains from residues
involved in contacts between F1-ATPase and IF1
have been drawn. For details of the color scheme
see the legend to Figure 1.
746
VOLUME 10 NUMBER 9 SEPTEMBER 2003 NATURE STRUCTURAL BIOLOGY
ARTICLES
higher than in the active enzyme23,24, and
ATP appears to be trapped in a nucleotide
binding site. It has been proposed that IF1
binds initially to F1-ATPase in an unproductive mode and upon hydrolysis of Mg-ATP,
this changes to a productive inhibitory complex, preventing product release11,25. In the
structure of the complex, the inhibited state
formed in the presence of ATP and crystallized in the presence of AMP-PNP traps ATP
or its analog in the βDP active site, which is
occupied by ADP in the reference state structure. Therefore, the structure seems to represent a pre-hydrolysis state in the catalytic
cycle. A number of studies have indicated that
ultimately ADP and IF1 are associated with a
final ‘dead-end’ product, not in the catalytic
cycle of F1-ATPase24,25. It remains possible
c
that if the crystals of the preformed complex
were grown in the presence of ATP (and not a
nonhydrolyzable analog), then ATP subsequently becomes hydrolyzed and a ‘dead-end’
complex with ADP in the inhibitory site
might be observed.
In principle, inhibition of hydrolysis could
be achieved either by preventing the conformational changes in the α and β subunits that
are required for rotary catalysis (and are
responsible for generating the rotation of the
γ subunit), or alternatively by directly
blocking the rotation of the γ subunit itself.
Figure 4 Conformational changes in the F1–IF1 complex relative to the reference state of F1-ATPase.
The
extent of the interaction between IF1 and
(a) Ribbon representation of superimposed βDP and γ subunits from the F1–IF1 complex (red) and the
the αDP and βDP subunits compared with the
reference state structure (blue). (b) Ribbon representation of superimposed αDP and γ subunits from
rather limited interactions with the γ subunit
the F1–IF1 complex (red) and the F1 reference state structure (blue); residues αDP405–408 are not
present in the F1–IF1 complex. (c) Stereo view of residues γ1–30 and γ221–272 and the C-terminal
strongly suggests that the former have the
domains of the (αβ)3 hexamer in the reference state structure (blue) and in the F1–IF1 complex (red),
dominant function in inhibition. A possible
obtained by superimposition of the N-terminal domains.
scheme for inhibition of hydrolysis by IF1
based on the binding change mechanism26 is
presented (Fig. 6). The ground state
8.7° and 5.8°, respectively (Fig. 4) about axes that lie close to the pep- (Fig. 6, step 1) binds ATP and IF1 at an open site (O), resulting in the
tides linking the nucleotide binding and C-terminal domains in the committed hydrolysis of ATP at an adjacent tight site, T* (Fig. 6,
respective subunits, in an orientation almost orthogonal to the axis of step 2). It is probable that ATP binds first, inducing a conformational
the (αβ)3 assembly. In consequence, the interface between the two change (O ➞ L′), which creates the binding site for IF1. ADP and
subunits is sufficiently open to accommodate the binding of IF1 (Fig. phosphate are released, and a cyclic interconversion of states occurs
5), resulting in an arrangement closely resembling the αTP-βTP inter- (Fig. 6, step 3). A second molecule of ATP binds to the newly formed
face in the reference state structure. In the other subunits, the r.m.s. open site (Fig. 6, step 4), resulting in another round of hydrolysis and
deviations in this domain are comparatively small (∼1.5 Å), with the product release (Fig. 6, step 5). The presence of IF1 bound to the loose
exception of the αE subunit (r.m.s. deviation 2.1 Å), which rotates by site, L (Fig. 6, step 4) prevents binding of a second inhibitor molecule.
3.4°, probably as a consequence of the changes in the adjacent subunits Also, the inhibitor blocks the conversion of the L site to a catalytically
(βDP and βE). Residue αE355 is also involved in contacts with residues competent tight site, even if ATP binds to the vacant open site
4 and 5 of IF1 (Fig. 3), and this interaction could explain the higher (Fig. 6, step 6). Further catalysis, which depends on a cyclic interconr.m.s. deviation found in that subunit (note that although residues 1–3 version of states, cannot occur. The crystal structure corresponds to
in IF1 are not present in the model, they are probably involved in addi- the state shown in step 5 in this scheme (Fig. 6). An alternative scheme
tional contacts with subunit αE). The rotation of the C-terminal is possible if IF1 binds to an L site rather than a closing (O ➞ L′) site. If
domain of the αDP subunit results in a slight opening of the nucleotide IF1 binds to site L in step 2 (Fig. 6), then after hydrolysis and product
binding pocket, which is probably responsible for the disorder release from the T* site, it will block the conversion of the L site to a T
observed in the adenine ring of the nucleotide bound to that subunit. site, effectively locking the complex in the state shown in step 3
(Fig. 6). Binding of IF1 prevents the closure of the αDP-βDP catalytic
DISCUSSION
interface, so that it resembles the αTP-βTP interface of the reference
The binding of IF1 to F1-ATPase requires an active enzyme and the state structure. By analogy with that structure, this may prevent the
presence of ATP22. In the inhibited complex, the ratio of ATP to ADP is close approach of the guanidino group of αDPArg373 to the β- and
© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
a
b
NATURE STRUCTURAL BIOLOGY VOLUME 10 NUMBER 9 SEPTEMBER 2003
747
© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
ARTICLES
a
b
c
d
e
f
Figure 5 The catalytic interface between subunits αDP and βDP in the reference state structure of F1-ATPase and in the F1–IF1 complex. (a) Space-filling
representations of the αDP and βDP subunits in the reference state structure. (b) Space-filling representations of the αDP and βDP subunits in the F1–IF1
complex with IF1 removed. (c) Space-filling representations of the αDP and βDP subunits in the F1–IF1 complex with IF1 present. (d–f) Ribbon representations
of the C-terminal domains of α and β subunits and part of the central coiled coil made from residues γ3–30 and γ221–272 viewed from underneath (90°
rotation) relative to the views in a–c. For details of colors, see the legend to Figure 1.
γ-phosphate groups of the nucleotide bound to the βDP subunit,
although the electron density for this side chain is too weak to confirm
this fully. This interpretation is consistent with the presence of ATP
bound to the βDP subunit, because αArg373 is known to be essential
for catalysis3,27.
Bovine IF1 is an equally effective inhibitor of both the bovine and
yeast enzymes28. Of the 32 residues in the α, β and γ subunits of bovine
F1-ATPase that are involved in binding to IF1, 25 of them are identical
in yeast F1-ATPase. A total of 21 of the 32 IF1 binding residues are also
identical in the enzyme from Escherichia coli, but neither the bovine
nor the yeast IF1 inhibits the bacterial enzyme (E.C. and J.E.W.,
unpublished observations). Bacteria have no IF1 homologs to regulate
the activity of their ATP synthases, and there is no physiological
requirement to inhibit the ATP hydrolytic activity of ATP synthase
under anaerobic conditions. Instead, as the PMF diminishes during
the onset of anaerobiosis, ATP made by substrate-level phosphorylation is hydrolyzed and provides energy to generate the PMF by
pumping protons out of the bacterial cytoplasm.
Nonetheless, the ε subunit of E. coli F1-ATPase does inhibit ATP
hydrolysis in vitro29, and a electron density map of the bacterial
enzyme30 at a resolution of 4.4 Å has been interpreted with the aid of a
structure of a complex of the ε subunit and residues 11–258 of the
γ subunit31 as showing a direct interaction between the C-terminal
α-helices of the ε subunit and the (αβ)3 hexamer32, supporting
crosslinking experiments33,34. Such an interaction might impede the
inhibitory effect of IF1 on the bacterial enzyme. These interpretations
of the structure of the E. coli F1-ATPase have led to the suggestion that
by interacting with the α and β subunits via its C-terminal helix, the
ε subunit might act in the intact ATP synthase as a ratchet to restrict or
prevent the conformational changes required for ATP hydrolysis, with
little effect on ATP synthesis33. There is no evidence that the
748
mitochondrial δ subunit (the equivalent of the bacterial ε subunit)
forms similar interactions7,8, and, given the extensive interactions of
this subunit with the ring of c subunits in the membrane domain7, augmented by the bovine ε subunit8, it is unlikely to do so. It has been suggested that in the mitochondrial enzyme IF1 fulfils a similar role to that
proposed for the bacterial ε subunit33. However, the mode of binding of
IF1 to bovine F1-ATPase does not support the notion that it acts as a
ratchet that could inhibit rotation in one direction (corresponding to
hydrolysis) in preference to the opposite direction (corresponding to
synthesis). Moreover, the interface (1,500 Å2) between F1 and IF1 is
Figure 6 The inhibition of the ATP hydrolytic activity of ATP synthase by IF1.
The scheme is based on the binding change mechanism of ATP hydrolysis26.
See text for further details.
VOLUME 10 NUMBER 9 SEPTEMBER 2003 NATURE STRUCTURAL BIOLOGY
ARTICLES
Table 1 Structural statistics for the bovine F1–IF1 complex
Data Collection
Space group
P212121
Unit cell dimensions (Å)
a
272.3
b
107.2
c
152.4
© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
Resolution (Å)
2.8
Number of reflections
109,367
Rmergea,b
0.081 (0.386)
<I / σ(I)>
5 (1.9)
Completenessa (%)
99.4 (97.3)
Multiplicitya
4.3 (3.5)
R-factora (%)
23.0 (44.0)
Rfreea,c (%)
27.7 (52.4)
R.m.s. deviations
Bonds (Å)
0.008
Angles (°)
1.4
aThe statistics for the highest-resolution shell are given in parentheses. bR
merge = Σ Σ |(Ihi –
<Ih>)| / ΣΣIhi where <Ih> is the weighted mean intensity for all observations of reflection h.
cR
was
calculated
for
5,507
reflections,
which
were
excluded
from
the
refinement.
free
much more extensive than the one between the bacterial ε subunit and
the C-terminal domain of a β-subunit found in the E. coli structure
(PDB entry 1JNV). Although both interfaces have some β-subunit
residues in common, the bacterial ε-subunit interacts with the βTP subunit and not the αDP-βDP interface as in the bovine complex.
IF1 inhibits its physiological target, ATP synthase, more effectively
than it inhibits F1-ATPase12. It is possible that binding of IF1 to ATP
synthase is enhanced by its C-terminal region making additional contacts with the peripheral stalk and/or with supernumerary subunits in
the Fo domain, but it is likely that the predominant interaction is with
the αDP-βDP interface in the F1 domain. The mechanism of reversal of
inhibition of ATP hydrolysis by IF1 in the ATP synthase in the presence
of a PMF remains a matter for speculation. Possible mechanisms are
that reversal of inhibition is promoted by structural changes induced
in the γ subunit by the PMF, and/or that its dissociation is driven by
formation of inactive higher oligomers of dimeric IF1 (ref. 16).
METHODS
Purification and crystallization of the F1–IF1 complex. The bovine F1–IF1
complex was prepared in the presence of ATP-Mg as described15. The purified
complex was stored as an ammonium sulfate precipitate. It was redissolved in
buffer (100 mM PIPES-NaOH, pH 6.6, 40 mM MgSO4, 0.04% (w/v) sodium
azide, 10% (v/v) glycerol and 0.002% (w/v) PMSF) and desalted on a PD-10
column (Amersham Biosciences). The protein solution was concentrated to
20 mg ml–1 at 20 °C with a centrifugal concentrator with a 100-kDa molecular
mass cutoff (Pall Filtron).
Crystals were grown from freshly desalted protein in microbatch plates
(Nalgene Nunc International). The drops made in D2O contained a solution
(2 µl) of 10 mM AMP-PNP, 300 mM NaCl, 16% (w/v) PEG 4000 and 5 mM
spermidine, and an equal volume of protein (2 µl; final concentration 10 mg
ml–1). Plates were covered with filtered liquid paraffin (∼8–10 ml; BDH
Laboratory Supplies) and incubated at 23 °C. Small crystals formed in 2 h, and
their growth was complete after 1 week. Typical final dimensions of crystals
were 400 × 200 × 80 µm.
Cryoprotection and data collection. Crystals were cryoprotected with 25%
(v/v) ethylene glycol. They were harvested with a cryoloop (Hampton
Research), plunged into liquid nitrogen and stored at 100 K. Data to a resolution of 2.8 Å were collected at 100 K on a ADSC CCD detector on beamline
ID14-4 (λ = 0.9366 Å) at the European Synchrotron Radiation Facility
(Grenoble, France).
Structure determination and refinement. The diffraction images were
indexed on an orthorhombic cell (space group P212121) with unit cell dimensions a = 272.3 Å, b = 107.2 Å, c = 152.4 Å, which are similar to those of native
F1-ATPase (a = 280.8 Å, b = 107.4 Å, c = 139.6 Å). The diffraction data were
integrated with MOSFLM35 and processed further with programs from the
CCP4 suite36. A close examination of the diffraction images revealed the
presence of semicontinuous diffraction, lying between layer lines predicted for
the orthorhombic unit cell, suggesting the presence of statistical disorder in
the crystals.
The structure was solved by molecular replacement with AmoRe37 using the
frozen native structure38 (PDB entry 1E1Q) as a starting model. After molecular replacement, the final R-factor and correlation coefficient were 43.7% and
58.0%, respectively, for data between resolutions 20 Å and 4 Å. Rigid-body
refinement and restrained refinement against a maximum-likelihood target
was carried out with CNS39. Electron density maps calculated at this stage of
the refinement (R = 32.9%, Rfree = 37.1%) were difficult to interpret. The electron density for βDP, βTP and the three α subunits was reasonably well defined,
with clear side chain densities, but the density for the βE and γ subunits was
poor, with frequent breaks. Additional uninterpreted density was also present.
A close examination of the electron density for the C-terminal domain of the
βE subunit revealed additional weak density, suggesting an alternative conformation for this domain. Fitting a Cα model to this density resulted in a conformation very similar to that of subunits βDP or βTP. This suggested that the
electron density represents an average of two structures of F1-ATPase, related by
a rotation of 120° such that, to a first approximation, the α subunits of one
F1-ATPase molecule would superimpose on different α subunits of the second
and similarly for the β subunits. This interpretation is consistent with the poor
density for the βE and γ subunits, which deviate most from the pseudo threefold symmetry of the molecule, and also the diffuse scattering on the diffraction
images.
Two models of the disorder were considered, corresponding to rotations of
+120° and –120°, respectively. After rigid-body refinement using a maximumlikelihood target, these models gave Rfree values of 36.7% and 33.3%, respectively, for data to a resolution of 2.8 Å, compared with 38.2% for a single
F1-ATPase. This provided strong support for the proposed disordered model
with a rotation of –120°. All subsequent refinement was carried out with two
molecules per asymmetric unit. Each molecule was assigned an occupancy of
0.5, and the second complete molecule was treated as an alternative conformation of the first in CNS. The presence of two molecules in the asymmetric unit
doubles the number of refined parameters, resulting in a poor ratio of data to
parameters (0.6:1). To help stabilize the refinement, noncrystallographic symmetry (NCS) restraints were applied to make the two molecules as similar as
possible, using the Rfree as the criterion to set the NCS weighting. Further
refinement and modeling of MgATP in the βDP subunit and IF1 residues 4–40 at
the interface between the αDP and βDP subunits, reduced the Rfree and working
R values to 27.7% and 23.0%, respectively. Averaging the density of the two
copies of the inhibitor made it much easier to fit the sequence. No solvent
molecules were included in the model. The final Rfree value is typical for a structure determined at this resolution, suggesting that the refined model is reliable.
However, it is certainly the case that the accuracy of the model, particularly the
side chains, will be somewhat lower than a typical structure determined at a
resolution of 2.8 Å. The stereochemistry was assessed with PROCHECK40,
which assigned 83.8% of the residues to the most favored region of the
Ramachandran plot, 15.4% to additional allowed regions and 0.8% to generously allowed regions. Structure comparisons and the determination of rotation axis parameters were carried out with the CCP4 program SUPERPOSE.
Interactions between F1-ATPase and IF1 were analyzed using the program
AREAIMOL, from the CCP4 suite. Figures 1–5 were generated with
BobScript41.
Coordinates. Coordinates and structure factors have been deposited in the
Protein Data Bank (accession code 1OHH).
ACKNOWLEDGMENTS
We thank the staff of beamline ID14 at European Synchrotron Radiation Facility
(ESRF), Grenoble, for help with data collection. E.C. was supported during part of
this work by a European Molecular Biology Organization Fellowship and by a
TMR Marie Curie Research Training Grant from the European Community.
NATURE STRUCTURAL BIOLOGY VOLUME 10 NUMBER 9 SEPTEMBER 2003
749
ARTICLES
COMPETING INTERESTS STATEMENT
The authors declare that they have no competing financial interests.
© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
Received 25 February; accepted 22 July 2003
Published online at http://www.nature.com/naturestructuralbiology/
1. Boyer, P.D. The ATP synthase: a splendid molecular machine. Annu. Rev. Biochem.
66, 717–749 (1997).
2. Walker, J.E. ATP synthesis by rotary catalysis (Nobel lecture). Angew. Chem. Int. Edn.
Engl. 37, 2309–2319 (1998).
3. Senior, A.E., Nadanaciva, S. & Weber, J. The molecular mechanism of ATP synthesis
by F1Fo-ATP synthase. Biochim. Biophys. Acta 1553, 188–211 (2002).
4. Collinson, I.R. et al. ATP synthase from bovine heart mitochondria. In vitro assembly
of a stalk complex in the presence of F1-ATPase and in its absence. J. Mol. Biol. 242,
408–421 (1994).
5. Karrasch, S. & Walker, J.E. Novel features in the structure of bovine ATP synthase.
J. Mol. Biol. 290, 379–384 (1999).
6. Rubinstein, J. & Walker, J.E. ATP synthase from Saccharomyces cerevisiae: location
of the OSCP subunit in the peripheral stalk region. J. Mol. Biol. 321, 613–619
(2002).
7. Stock, D., Leslie, A.G.W. & Walker, J.E. Molecular architecture of the rotary motor in
ATP synthase. Science 286, 1700–1705 (1999).
8. Gibbons, C., Montgomery, M.G., Leslie, A.G.W. & Walker, J.E. The structure of the
central stalk in bovine F1-ATPase at 2.4 Å resolution. Nat. Struct. Biol. 7,
1055–1061 (2000).
9. Abrahams, J.P., Leslie, A.G.W., Lutter, R. & Walker, J.E. Structure at 2.8 Å resolution
of F1-ATPase from bovine heart mitochondria. Nature 370, 621–628 (1994).
10. Pullman, M.E. & Monroy, G.C. A soluble heat stable protein in mitochondria from
bovine heart that inhibits ATP hydrolase activity. J. Biol. Chem. 238, 3762–3769
(1963).
11. Green, D.W. & Grover, G.J. The IF1 inhibitor protein of the mitochondrial F1FoATPase. Biochim. Biophys. Acta 1458, 343–355 (2000).
12. van Raaij, M.J. et al. The ATPase inhibitor protein from bovine heart mitochondria:
the minimal inhibitory sequence. Biochemistry 35, 15618–15625 (1996).
13. Gordon-Smith, D.J. et al. Solution structure of a C-terminal coiled-coil domain from
bovine IF1: the inhibitor protein of F1-ATPase. J. Mol. Biol. 308, 325–339 (2001).
14. Cabezon, E., Runswick, M.J., Leslie, A.G.W. & Walker, J.E. The structure of bovine
IF1, the regulatory subunit of mitochondrial F1-ATPase. EMBO J. 20, 6990–6996
(2001).
15. Cabezon, E., Arechaga, I., Butler, P.J.G. & Walker, J.E. Dimerization of bovine F1ATPase by binding the inhibitor protein, IF1. J. Biol. Chem. 275, 28353–28355
(2000).
16. Cabezon, E., Butler, P.J.G., Runswick, M.J. & Walker, J.E. Modulation of the
oligomerization state of the bovine F1-ATPase inhibitor protein, IF1, by pH. J. Biol.
Chem. 275, 25460–25464 (2000).
17. Rouslin, W. Protonic inhibition of the mitochondrial oligomycin-sensitive adenosine
5′-triphosphatase in ischemic and autolyzing cardiac muscle. Possible mechanism
for the mitigation of ATP hydrolysis under nonenergizing conditions. J. Biol. Chem.
258, 9657–9661 (1983).
18. Rouslin, W. Factors affecting the reactivation of the oligomycin-sensitive adenosine
5′-triphosphatase and the release of ATPase inhibitor protein during the re-energization of intact mitochondria from ischemic cardiac muscle. J. Biol. Chem. 262,
3472–3476 (1987).
19. Rouslin, W. & Broge, C.W. Regulation of mitochondrial matrix pH and adenosine
5′-triphosphatase activity during ischemia in slow heart-rate hearts. Role of Pi/H+
symport. J. Biol. Chem. 264, 15224–15229 (1989).
20. Mimura, H., Hashimoto, T., Yoshida, Y., Ichikawa, N. & Tagawa, K. Binding of an
intrinsic ATPase inhibitor to the interface between α- and β-subunits of F1Fo-ATPase
750
upon de-energization of mitochondria. J. Biochem. (Tokyo) 113, 350–354 (1993).
21. Jackson, P.J. & Harris, D.A. The mitochondrial ATP synthase inhibitor protein binds
near the C terminus of the F1 β-subunit. FEBS Lett. 229, 224–228 (1988).
22. Power, J., Cross, R.L. & Harris, D.A. Interaction of F1-ATPase, from ox heart mitochondria with its naturally occurring inhibitor protein. Studies using radio-iodinated
inhibitor protein. Biochim. Biophys. Acta 724, 128–141 (1983).
23. Milgrom, Y.M. When beef-heart mitochondrial F1-ATPase is inhibited by inhibitor protein a nucleotide is trapped in one of the catalytic sites. Eur. J. Biochem. 200,
789–795 (1991).
24. Penin, F., Di Pietro, A., Godinot, C. & Gautheron, D.C. Fate of nucleotides bound to
reconstituted Fo-F1 during adenosine 5′-triphosphate synthesis activation or hydrolysis: role of protein inhibitor and hysteretic inhibition. Biochemistry 27, 8969–8974
(1988).
25. Pedersen, P.L., Schwerzmann, K. & Cintron, N.M. Regulation of the synthesis and
hydrolysis of ATP in biological systems: role of peptide inhibitors of H+-ATPases. Curr.
Top. Bioenerg. 11, 149–199 (1981).
26. Boyer, P.D. The binding change mechanism for ATP synthase: some probabilities and
possibilities. Biochim. Biophys. Acta 1140, 215–250 (1993).
27. Nadanaciva, S., Weber, J., Wilke-Mounts, S. & Senior, A.E. Importance of F1-ATPase
residue α-Arg376 for catalytic transition state stabilisation. Biochemistry 38,
15493–15499 (1999).
28. Cabezon, E., Butler, P.J, Runswick, M.J., Carbajo, R.J. & Walker, J.E. Homologous
and heterologous inhibitory effects of ATPase inhibitor proteins on F1-ATPases.
J. Biol. Chem. 277, 41334–41341 (2002).
29. Klionsky, D.J., Bruslow, W.S.A. & Simoni, R.D. In vivo evidence for the role of the
ε-subunit as an inhibitor of the proton-translocating ATPase of Escherichia coli.
J. Bacteriol. 160, 1055–1060 (1985).
30. Hausrath, A.C., Gruber, G., Matthews, B.W. & Capaldi, R.A. Structural features of the
γ-subunit of the Escherichia coli F1-ATPase revealed by a 4.4 Å resolution map
obtained by X-ray crystallography. Proc. Natl. Acad. Sci. USA 96, 13697–13702
(1999).
31. Rodgers, A.J. & Wilce, M.C. Structure of the γ-ε complex of ATP synthase. Nat. Struct.
Biol. 7, 1051–1054 (2000).
32. Hausrath, A.C., Capaldi, R.A. & Matthews, B.W. The conformation of the ε- and γ-subunits within the Escherichia coli F1-ATPase. J. Biol. Chem. 276, 47227–47232
(2001).
33. Tsunoda, S.P. et al. Large conformational changes of the ε-subunit in the bacterial
F1Fo ATP synthase provide a ratchet action to regulate this rotary motor enzyme.
Proc. Natl. Acad. Sci. USA 98, 6560–6564 (2001).
34. Aggeler, R., Haughton, M.A. & Capaldi, R.A. Disulfide bond formation between the
COOH-terminal domain of the β-subunits and the γ- and ε-subunits of the Escherichia
coli F1-ATPase. Structural implications and functional consequences. J. Biol. Chem.
270, 9185–9191 (1995).
35. Leslie, A.W.G. Joint CCP4 and ESF-EACMB Newsletter on Protein Crystallography
Vol. 26 (Daresbury Laboratory, Warrington, UK, 1992).
36. Collaborative Computational Project, Number 4. The CCP4 Suite: programs for
protein crystallography. Acta Crystallogr. D 50, 760–763 (1994).
37. Navaza, J. AmoRe: an automated package for molecular replacement.
Acta Crystallogr. A 50, 157–163 (1994).
38. Braig, K., Menz, R.I., Montgomery, M.G., Leslie, A.G.W. & Walker, J.E. Structure of
bovine mitochondrial F1-ATPase inhibited by Mg2+ ADP and aluminum fluoride.
Structure 8, 567–573 (2000).
39. Brunger, A.T. et al. Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. A 54, 905–921 (1998).
40. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr.
26, 283–291 (1993).
41. Esnouf, R.M. An extensively modified version of MolScript that includes greatly
enhanced coloring capabilities. J. Mol. Graph. 15, 132–134 (1997).
VOLUME 10 NUMBER 9 SEPTEMBER 2003 NATURE STRUCTURAL BIOLOGY