878
Biochemical Society Transactions (2005) Volume 33, part 4
Subunit movement in individual H+-ATP synthases
during ATP synthesis and hydrolysis revealed by
fluorescence resonance energy transfer
M. Börsch*1 and P. Gräber†
*3. Physikalisches Institut, Universität Stuttgart, Pfaffenwaldring 57, 70569 Stuttgart, Germany, and †Institut für Physikalische Chemie,
Albert-Ludwigs-Universität Freiburg, Albertstr. 23 a, 79104 Freiburg, Germany
Abstract
F-type H+ -ATP synthases synthesize ATP from ADP and phosphate using the energy supplied by a
transmembrane electrochemical potential difference of protons. Rotary subunit movements within the
enzyme drive catalysis in either an ATP hydrolysis or an ATP synthesis direction respectively. To monitor
these subunit movements and associated conformational changes in real time and with subnanometre
resolution, a single-molecule FRET (fluorescence resonance energy transfer) approach has been developed
using the double-labelled H+ -ATP synthase from Escherichia coli. After reconstitution into a liposome, this
enzyme was able to catalyse ATP synthesis when the membrane was energized.
Subunit movements in F1 domains
+
The H -ATP synthase from Escherichia coli is a large membrane-bound enzyme that consists of eight different subunits.
These subunits comprise two parts: the hydrophilic F1 part
with subunit composition α 3 β 3 γ δε containing three catalytic
nucleotide binding sites and the membrane-integrated Fo part
with subunits ab2 c10 containing the proton transport sites.
The enzyme catalyses the synthesis of ATP using the electrochemical potential difference of protons across the lipid
membrane and the reverse reaction, the proton transportcoupled ATP hydrolysis [1–3].
The α- and β-subunits are arranged in a barrel-like
hexameric structure around the central γ -subunit and catalysis is associated with a stepwise rotational motion of the
γ - and ε-subunits relative to α 3 β 3 . First direct experimental
evidence for a rotational mechanism of catalysis was provided
by biochemical methods in 1995 using cross-links between
the γ -subunit and the β-subunits specifically tagged with a
radioactive label [4]. Spectroscopic methods using the photoselection with polarized laser light were followed by the
single-molecule anisotropy measurements of the γ -subunit
rotation using immobilized F1 domains [5,6].
The most elegant way of a direct demonstration of γ -subunit rotation was the videomicroscopic experiment by Noji
et al. in 1997 [7]. The authors attached the α 3 β 3 γ fragment
from the bacterium PS3 to a glass surface through the N-terminal histidine tags at the three β-subunits and used a 3 µm
fluorescent actin filament bound to the γ -subunit as a pointer
of the actual γ -subunit orientation. Starting hydrolysis by the
Key words: ATP synthase, catalysis, fluorescence resonance energy transfer (FRET), hydrolysis,
single-molecule fluorescence, subunit rotation.
Abbreviations used: FRET, fluorescence resonance energy transfer; TMR, tetramethylrhodamine.
1
To whom correspondence should be addressed (email [email protected]).
C 2005
Biochemical Society
addition of ATP resulted in the counterclockwise rotation of
the actin filament, which was observed for several minutes.
This approach was later refined using smaller polymer beads
or magnetic beads. The original finding of a three-stepped
rotation of the γ -subunit has emerged into a more complex
picture with at least one intermediary angle-resolved substep
and the distinct stopping positions of the γ -subunit have
been correlated with the binding events of ATP, the catalytic
reaction of ATP hydrolysis and the sequential release of the
products ADP and phosphate [8].
The single-molecule FRET (fluorescence
resonance energy transfer) approach
with Fo F1
ATP synthesis requires that the enzyme with all subunits
is embedded into a lipid bilayer and, additionally, a proton
motive force has to be applied. This has been achieved by
reconstitution of the holoenzyme Fo F1 into the lipid vesicles.
The required pH difference pH and the electric potential
difference are generated by a rapid injection of the
proteoliposomes (equilibrated at low pH and low internal
K+ concentration) into a buffer at high pH and high K+
concentration in the presence of the K+ carrier valinomycin.
Depending on a variety of parameters, ATP synthesis can
be measured up to several minutes before pH and are
dissipated [9,10].
To monitor the intramolecular rotary motions of the γ -,
ε- or c-subunits upon ATP synthesis, the size for the markers
of rotation described above had to be reduced significantly.
Organic fluorophores have a diameter of approx. 1 nm. When
they are covalently bound to cysteine residues at appropriate
sites, they do not prevent the rotation of subunits γ or ε,
which have to pass through the space between the peripheral
Mechanisms of Bioenergetic Membrane Proteins
Figure 1 Model of the H+ -ATP synthase from E. coli according to [14]
(A) The membrane-embedded Fo -part is located at the bottom, the F1 -part is at the top. The rotating subunits are shown
in black, the static subunits are shown in grey. Two positions for fluorophore labelling have been introduced into ‘off-axis
domains’ of the rotor, utilizing a cysteine residue either in the γ -subunit (γ 106, [11–16]) or in the ε-subunit (ε56). On
the stator specific labelling either of a lysine residue on the β-subunits (β4, [13]) or of a cysteine residue at the dimeric
b-subunits (b64, [11,12,14,15,17]) was used for single-molecule FRET studies. The arrow indicates the direction of rotation
during ATP synthesis. (B) The FRET-labelled H+ -ATP synthase in a liposome is diffusing through the confocal volume of laser
excitation and photon detection.
stalk consisting of the b-subunit dimer and the central axis
of rotation (Figure 1). We attach one fluorophore to an
amino acid at the surface of either the γ - or ε-subunit and
choose positions with large distances towards the expected
axis of rotation in Fo F1 [11–17]. The radius of fluorophore
rotation is estimated from the structure to be in the range
of 3–3.5 nm. The rotation of γ or ε results in sequential
distance changes between these off-axis parts of the subunits
with respect to a fixed point outside the rotor. The shortest
and the longest distances towards a second position at the
fixed b-dimer are expected to be in the range 1–10 nm
(i.e. the diameter of the F1 part). The perfect spectroscopic
method to detect distance changes in this range is FRET.
FRET is the non-radiative process of energy transfer between
two molecules or fluorophores [18]. The directly excited
fluorophore is called the FRET donor, the second one is
called the FRET acceptor and it is in the electronic ground
state. Energy transfer from the donor to the acceptor results
in a decrease of fluorescence intensity of the donor and an
increase of fluorescence of the acceptor. The FRET efficiency
depends on the distance between the two fluorophores,
their spectral properties and the relative orientation of their
transition dipole moments. Using rhodamines like TMR
(tetramethylrhodamine) as the FRET donor and indodicarbocyanines like Cy5 as the FRET acceptor yields a 50%
FRET efficiency at a distance of R0 = 6–7 nm (‘Förster
radius’) and, accordingly, reliable distance measurements can
be made between 3 and 9 nm.
To carry out FRET measurements at the single-molecule
level, the exciting light must be confined to a very small
volume (a few femtolitres) to excite only one fluorophore
[19]. This is achieved by diffraction-limited focusing of a
laser beam with a microscope objective into the sample
buffer. The sample contains a very low concentration of
the labelled enzyme (100 pM), i.e. only 0.1 molecule is present in the focal volume on time average. When a proteoliposome reaches the laser focus during its diffusion (Figure 1B) the FRET donor fluorophore is frequently excited
(106 s−1 ). It relaxes either by emitting a photon (donor fluorescence) or by transferring the energy to the FRET acceptor
(acceptor fluorescence). Confocal detection with a pinhole in
the emission pathway rejects the out-of-focus fluorescence.
Efficient optical filters in combination with single photon
sensitivity of the avalanche photodiodes enable the observation of double-labelled enzymes during their diffusion
through the confocal volume. The fluorescence is spectrally
separated into two wavelength regions for separate detection
of donor and acceptor. The stochastic diffusion process
results in strong fluctuations of both fluorescence intensities.
However, the FRET efficiency, EFRET , i.e. the corrected ratio
C 2005
Biochemical Society
879
880
Biochemical Society Transactions (2005) Volume 33, part 4
between the acceptor fluorescence and the sum of both
acceptor and donor fluorescence, remains constant as long
as the intramolecular distance does not change.
Stepwise γ -subunit rotation in opposite
directions
After the introduction of a cysteine residue in the b-subunits
at position 64, we labelled it with TMR-maleimide as the
FRET donor. The Cys106 of the γ -subunit was labelled with
Cy5-monomaleimide as the FRET acceptor [11,12]. In
the presence of AMP-PNP (adenosine 5 -[β,γ -imido]-triphosphate), a non-hydrolysable ATP analogue, three distinct
FRET efficiencies were observed. The FRET states were
called L state (low FRET efficiency: small energy transfer due
to a large donor–acceptor distance), M state (medium FRET
efficiency: approx. 50% transfer efficiency) and H state (high
FRET efficiency: large energy transfer due to a short donor–
acceptor distance). These FRET efficiencies remained constant within a photon burst (∼ 100 ms), indicating a fixed
distance between the two labels and, therefore, the absence
of γ -subunit movements in this time range. In the presence of 1 mM ATP, fluctuations of the FRET efficiencies
were detected within the photon burst, indicating a stepwise
rotary motion of the γ -subunit during ATP hydrolysis. The
FRET efficiencies, i.e. the FRET levels, were comparable
with those in the presence of AMP-PNP; however, their
magnitude changed stepwise and sequentially. Fo F1 contains
two b-subunits, and therefore, the donor was bound at two
different sites, which have different distances to the acceptor. Unfortunately, we found two sequences of FRET
level transitions during ATP hydrolysis with similar probability, in the L-M-H-L order as well as in the H-ML-H order. Thus a correlation of one sequence with a unidirectional γ -subunit rotation during ATP hydrolysis was
not possible.
In the next experiment, we labelled one lysine in the βsubunit at position 4 (see Figure 1A) with a rhodamine as the
FRET donor and the γ -subunit at the position 106 with a
indodicarbocyanine as the FRET acceptor [13]. Both labels
were attached to the F1 part, before F1 was reassembled
with Fo in liposomes. During ATP hydrolysis, a preferred
sequence of FRET level transitions could be attributed to
the counterclockwise rotation of γ . Changing the conditions
to the ATP synthesis direction by generating pH plus in the presence of ADP and phosphate resulted in stepwise
FRET-level transitions in the opposite direction. However,
the signal-to-noise ratio had to be improved significantly.
Finally, we resolved the ambiguity of the localization on
the b-subunits as follows: the donor TMR was attached to
position γ 106 and the two b-subunits were cross-linked
with the acceptor Cy5-bismaleimide at the b64-position
(Figure 1A) [14,15]. With this system we found the FRET
level sequence H → M → L → H → during ATP hydrolysis
for 72% of the bursts (see Figure 2A) and the reversed
sequence L → M → H → L → during ATP synthesis for
83% of the burst (see Figure 2B). Independently, the FRET
C 2005
Biochemical Society
Figure 2 Photon bursts of single FRET-labelled H+ -ATP synthases
in liposomes with TMR as FRET donor at γ 106 and Cy5bismaleimide as FRET acceptor at b64 [14]
In the lower panels, the fluorescence intensity trajectories of the two
channels are shown (grey, TMR; black, Cy5). In the upper panels, the
calculated FRET efficiencies are depicted as circles and the mean value for
each level as a black line. (A) During ATP hydrolysis, the enzyme is detected at t = 60 ms in the high-FRET level; it switches then to the medium-FRET level and finally to the low-FRET level before leaving the
confocal detection volume. The photon burst at t = 230 ms is due to an
enzyme labelled only with TMR (‘donor only’). (B) During ATP synthesis,
the enzyme is detected at t = 230 ms in the L level; it switches then
to the M level and finally to the H level before it leaves the confocal volume. This is the opposite sequence of FRET levels as during ATP hydrolysis. The vertical lines indicate the beginning and end of the FRET levels.
levels and the order of transitions were analysed by measuring the fluorescence lifetime of the FRET donor in the photon
burst yielding the same result as the analysis of the fluorescence intensities.
Single-molecule dwell times and ensemble
turnover rates
During catalysis, several FRET levels were found within each
photon burst. The FRET level duration (‘dwell time’) was
measured and summarized in histograms. The first and the last
FRET levels have been omitted in the dwell time histograms
since their duration (before entering and after leaving the observation volume) remain unknown. Therefore only the
intermediary FRET levels of photon bursts with three and
Mechanisms of Bioenergetic Membrane Proteins
Figure 3 Dwell time histograms of the H+ -ATP synthases [14]
during catalysis
(A) During ATP hydrolysis, the intermediary FRET levels with three and
more FRET levels were binned into 5 ms intervals. (B) During ATP
synthesis, the intermediary FRET levels were binned into 15 ms intervals.
more FRET levels contribute to the dwell time histograms
shown in Figure 3. Depending on the time resolution of
the fluorescence intensity trajectories (currently 1 ms, see
Figure 2) we define a FRET level by a minimum duration of
4 ms and a maximum S.D. = 0.15 of the mean FRET efficiency
of this level. Shorter dwell times are not taken into account
and this is one reason for the many apparently ‘wrong’ FRET
level sequences found during catalysis.
What are the biochemical events leading to the stepped
rotation of the γ -subunit? Upon ATP hydrolysis at low ATP
concentrations the video microscopic experiments with single
F1 fragments revealed an 80–90◦ rotation of γ associated with
the binding of ATP to an empty catalytic site and a 40–30◦
rotation associated with the release of ADP and phosphate
[20]. In our single-molecule FRET approach, we used 1 mM
ATP, and therefore, the waiting time of the γ -subunit, i.e. the
duration of a constant FRET level (‘dwell time’), is due
to the hydrolysis reaction itself but not due to the delay
until the next ATP molecule is bound to the enzyme. The
mean dwell time τ hyd was calculated from a monoexponential
decay fit of the data in Figure 3(A) and yielded τ hyd = 19 ms.
The rate of ATP hydrolysis was calculated from biochemical
ensemble measurements to yield vhyd = 67 s−1 , giving a turnover time of τ hyd = 15 ms. Both rates are the same within
error limits. Actually, we expected the single-molecule rate
to be higher, since in this case only the active enzymes are
contributing to the rate, whereas in the ensemble measurements the rate refers to all enzymes, i.e. the active and the
inactive ones.
For ATP synthesis, this direct comparison between the
single-molecule dwell times and the ensemble turnover
is more complicated. The ensemble rates are determined
from the initial turnover immediately after generating the
pH and . It includes all enzymes, active and inactive
ones, and therefore, this rate should be smaller than the singlemolecule rate as discussed for ATP hydrolysis. On the other
hand, the single-molecule dwell times are collected between
30 s and 2 min after mixing. This time window was chosen
since the data can be collected only after the disappearance of
the turbidity changes due to mixing. However, dissipation
of pH and does not allow longer observation times.
Therefore the actual value of the proton motive force
experienced by the individual Fo F1 under observation is
unknown and thus the measured dwell times cannot be
correlated with the driving force for ATP synthesis. Despite
these problems the single-molecule ATP synthesis turnover
(τ s = 51 ms) is in reasonable agreement with the ensemble
turnover time (τ = 43 ms).
Some limitations of the single-molecule FRET approach
with freely diffusing liposomes have to be overcome in future:
(1) The observation time of a few hundred milliseconds is
too short to allow monitoring of several repeating rotations.
Possibly, this problem might be solved by immobilization
of the proteoliposomes [21,22]. (2) The small number of
detected photons per time interval limits the precision of the
distance measurements. (3) Quenching of the protein-bound
fluorophores like TMR [23] has to be avoided by the use
of novel fluorophores with higher quantum yield and better
photostability. This would allow further studies which give
a detailed view of structural dynamics and conformational
changes of a single H+ -ATP synthase during catalysis.
We thank M. Diez, B. Zimmermann, S. Steigmiller and J. Petersen
(Freiburg) and N. Zarrabi and M. Düser (Stuttgart) for their exceptional contributions to the development and success of the singlemolecule FRET investigations of H+ -ATP synthases; C.A.M. Seidel
(Düsseldorf) for the help with the design of our confocal singlemolecule FRET set-up and the gift of the Cy5-bismaleimide
fluorophore and W. Wangler (Freiburg) for excellent technical
assistance. Parts of this work have been supported by the
Landesstiftung Baden-Württemberg in the network of competence
‘Functional Nanodevices’.
References
1
2
3
4
Boyer, P. (1997) Annu. Rev. Biochem. 66, 717–749
Capaldi, R.A. and Aggeler, R. (2002) Trends Biochem. Sci. 27, 154–160
Weber, J. and Senior, A.E. (2003) FEBS Lett. 545, 61–70
Duncan, T.M., Bulygin, V.V., Zhou, Y., Hutcheon, M.L. and Cross, R.L.
(1995) Proc. Natl. Acad. Sci. U.S.A. 92, 10964–10968
5 Sabbert, D., Engelbrecht, S. and Junge, W. (1996) Nature (London) 381,
623–625
C 2005
Biochemical Society
881
882
Biochemical Society Transactions (2005) Volume 33, part 4
6 Häsler, K., Engelbrecht, S. and Junge, W. (1998) FEBS Lett. 426, 301–304
7 Noji, H., Yasuda, R., Yoshida, M. and Kinosita, Jr, K. (1997)
Nature (London) 386, 299–302
8 Yasuda, R., Noji, H., Yoshida, M., Kinosita, Jr, K. and Itoh, H. (2001)
Nature (London) 410, 898–904
9 Jagendorf, A.T. and Uribe, E. (1966) Proc. Natl. Acad. Sci. U.S.A. 55,
170–177
10 Fischer, S. and Gräber, P. (1999) FEBS Lett. 457, 327–332
11 Börsch, M., Diez, M., Zimmermann, B., Reuter, R. and Gräber, P. (2002)
in Fluorescence Spectroscopy, Imaging and Probes. New Tools in
Chemical, Physical and Life Sciences (Kraayenhof, R., Visser, A.J.W.G. and
Gerritsen, H.C., eds.), pp. 197–207, Springer-Verlag, Heidelberg
12 Börsch, M., Diez, M., Zimmermann, B., Reuter, R. and Gräber, P. (2002)
FEBS Lett. 527, 147–152
13 Börsch, M., Diez, M., Zimmermann, B., Trost, M., Steigmiller, S. and
Gräber, P. (2003) Proc. SPIE Int. Soc. Opt. Eng. 4962, 11–21
14 Diez, M., Zimmermann, B., Börsch, M., König, M., Schweinberger, E.,
Steigmiller, S., Reuter, R., Felekyan, S., Kudryavtsev, V., Seidel, C.A.M.
et al. (2004) Nat. Struct. Mol. Biol. 11, 135–141
C 2005
Biochemical Society
15 Boldt, F.-M., Heinze, J., Diez, M., Petersen, J. and Börsch, M. (2004)
Anal. Chem. 76, 3473–3481
16 Steigmiller, S., Zimmermann, B., Diez, M., Börsch, M. and Gräber, P.
(2004) Bioelectrochemistry 63, 79–85
17 Zarrabi, N., Zimmermann, B., Diez, M., Gräber, P., Wrachtrup, J. and
Börsch, M. (2005) Proc. SPIE Int. Soc. Opt. Eng. 5699, 175–188
18 Förster, T. (1948) Ann. Phys. 2, 55–70
19 Ha, T. (2001) Methods 25, 78–86
20 Yasuda, R., Noji, H., Yoshida, M., Kinosita, K. and Itoh, H. (2001)
Nature (London) 410, 898–904
21 Stamou, D., Duschl, C. and Vogel, H. (2003) Angew. Chem. Int. Ed. Engl.
42, 5580–5583
22 Luo, T.-J.M., Soong, R., Lan, E., Dunn, B. and Montamagno, C. (2005)
Nat. Mater. 4, 220–224
23 Börsch, M., Turina, P., Eggeling, C., Fries, J.R., Seidel, C.A., Labahn, A. and
Gräber, P. (1998) FEBS Lett. 437, 251–254
Received 13 April 2005