26. G. Kaim, F. Wehrle, U. Gerike, P. Dimroth, Biochemistry
36, 9185 (1997).
27. T. Murata et al., J. Biol. Chem. 278, 21162 (2003).
28. H. Arai et al., J. Biol. Chem. 263, 8796 (1999).
29. A. Holzenburg et al., Eur. J. Biochem. 213, 21 (1993).
30. R. W. Hendrix, Proc. Natl. Acad. Sci. U.S.A. 75, 4779
(1978).
31. D. R. Thomas, D. G. Morgan, D. J. DeRosier, Proc. Natl.
Acad. Sci. U.S.A. 96, 10134 (1999).
32. Single-letter abbreviations for the amino acid residues
are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G,
Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro;
Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
33. J. D. Thompson, D. G. Higgins, T. J. Gibson, Nucleic
Acids Res. 22, 4673 (1994).
34. This work was supported by the Medical Research Coun-
cil. T.M. was supported partly by a Fellowship for Re-
search Abroad from the Japan Society for the Promotion
of Science. We thank the staff at the European Synchro-
tron Radiation Facility in Grenoble for their help. Coor-
dinates and structure factors (accession codes 2bl2 and
r2bl2sf) have been deposited in the Protein Data Bank.
Supporting Online Material
www.sciencemag.org/cgi/content/full/1110064/DC1
Materials and Methods
SOM Text
Fig. S1
References
21 January 2005; accepted 8 March 2005
Published online 31 March 2005;
10.1126/science.1110064
Include this information when citing this paper.
Structure of the Rotor Ring of
F-Type Naþ-ATPase from
Ilyobacter tartaricus
Thomas Meier,1 Patrick Polzer,2 Kay Diederichs,2*
Wolfram Welte,2 Peter Dimroth1*
In the crystal structure of the membrane-embedded rotor ring of the sodium
ion–translocating adenosine 5¶-triphosphate (ATP) synthase of Ilyobacter
tartaricus at 2.4 angstrom resolution, 11 c subunits are assembled into an
hourglass-shaped cylinder with 11-fold symmetry. Sodium ions are bound in a
locked conformation close to the outer surface of the cylinder near the middle
of the membrane. The structure supports an ion-translocation mechanism in
the intact ATP synthase in which the binding site converts from the locked
conformation into one that opens toward subunit a as the rotor ring moves
through the subunit a/c interface.
In the F1Fo ATP synthase, the cytoplasmic F1
catalytic domain (subunits a3b3gde) is linked
by means of a central and a peripheral stalk
(subunits g/e and b2/d, respectively) to the in-
trinsic membrane domain called Fo (subunits
ab2c10-14). Each of these domains functions as
a reversible rotary motor and exchanges energy
with the opposite motor by mechanical rotation
of the central stalk. During ATP synthesis,
energy stored in an electrochemical gradient of
protons or Naþ ions fuels the Fo motor, which
causes the stalk to rotate with the inherently
asymmetric g subunit acting as a camshaft to
continuously change the conformation of each
catalytic b subunit. These sequential intercon-
versions, which result in ATP synthesis, en-
dow the binding sites with different nucleotide
affinities Efor reviews see (1–3)^. The rotational
model, which explains a wealth of bio-
chemical and kinetic data, is impressively sup-
ported by the crystal structure of F1 (4) and
was experimentally verified by biochemical,
spectroscopic, and microscopic techniques (5).
The Fo motor consists of an oligomeric ring
of c subunits that is abutted laterally by the a
and b2 subunits (6). The c ring, together with
g/e subunits, forms the rotor assembly, which
spins against the stator components ab2da3b3.
Ion translocation at the interface between sub-
unit a and the c ring, driven by the ion motive
force, is thought to generate torque (7–10)
applied to the g subunit, which is then used to
promote the conformational changes required
for ATP synthesis at the F1 catalytic sites.
Despite intense efforts, little is known
about the structural details of Fo. This lack of
information hinders our understanding of
how this molecular motor functions. The
nuclear magnetic resonance (NMR) struc-
tures of the c monomer of Escherichia coli
(7, 11) showed that the protein is folded into
two a helices linked by a loop. Other struc-
tural studies indicated that the c subunits of
the oligomer are tightly packed into two con-
centric rings of helices (12, 13). The number
of c monomers per ring varies between n 0
10, 11, and 14 units in the ATP synthases
from yeast (12), I. tartaricus (13), and
spinach chloroplasts (14), respectively.
Structure of the c ring. We chose to de-
termine the crystal structure of the I. tartaricus
c ring because of its inherent stability and rel-
ative ease of isolation (15). After purification
and crystallization of wild-type c ring (16), the
structure was solved (table S1) by molecular
replacement (17) using a medium-resolution
(6 )) c-ring backbone model derived from
electron crystallography (13). The asymmetric
unit of the crystal contains 4 c rings. These
rings are arranged in two parallel, but laterally
translated, c-ring dimers each formed by a
coaxial association of two rings that interact
with their termini in a tail-to-tail fashion. A non-
crystallographic symmetry restraint was imposed
during refinement at 2.4 ) over the 44 mono-
mers in the asymmetric unit (3916 residues)
with the exclusion of the loop regions that form
crystal contacts, which substantially improved
the electron density and resulted in an atomic
model without Ramachandran plot outliers.
The electron density map of a single c ring
shows a cylindrical, hourglass-shaped protein
complex of È70 ) in height, and with an outer
diameter of È40 ) in the middle and È50 ) at
the top and bottom. Eleven c subunits, each
composed of two membrane-spanning a helices
forming a hairpin, are arranged around an 11-
fold axis, creating a tightly packed inner ring
with their N-terminal helices (Fig. 1). The C-
terminal helices pack into the grooves formed
between N-terminal helices, producing an outer
ring, in agreement with previous medium-
resolution structures (12, 13). In the electron
density map, the backbone and side chains of
all amino acids are clearly defined, except for
the C-terminal glycine. The N- and C-terminal
helices are connected by a loop formed by
the highly conserved peptide Arg45, Gln46,
and Pro47, which is exposed to the cyto-
plasmic surface (Fig. 2) (18, 19). The chain
termini are exposed to the periplasm.
The C-terminal helices are shorter than the
N-terminal helices, owing to a break at Tyr80
followed by another short helix of one turn
(Figs. 1 and 2). For each helix, an individual
plane can be found that roughly contains the
axis of the helix and the c-ring symmetry axis
(Fig. 1A). All helices show a bend of about 20-
in the middle of the membrane (at Pro28 and
Glu65 in the N-terminal and the C-terminal
helices, respectively), causing the narrow part
of the hourglass shape. Moreover, the bend
tilts the helices in the cytoplasmic half out of
the plane by È10-, yielding a right-handed
twisted packing (Fig. 1A). When the c ring is
viewed from the cytoplasm, it rotates counter-
clockwise during ATP synthesis (20) against
the drag imposed by the F1 motor components.
Thus, the resulting torque might decrease the
bend and increase the interhelical distance in
the cytoplasmic part of the c ring, depending
on the energies involved. Such a conforma-
tional change under load might serve to store
elastic energy in the c ring, adding to that de-
scribed for the central and peripheral stalk
subunits (21). A change in the twist of the
helices is supported by calculations (22) that
show in the lowest-order mode a torsional
1Institut fu¨r Mikrobiologie, Eidgeno¨ssische Technische
Hochschule (ETH), Zu¨rich Ho¨nggerberg, Wolfgang-Pauli-
Str. 10, CH-8093 Zu¨rich, Switzerland. 2Fachbereich Biologie,
Universita¨t Konstanz M656, D-78457 Konstanz, Germany.
*To whom correspondence should be addressed.
E-mail: dimroth@micro.biol.ethz.ch (P.D.); kay.diederichs@
uni-konstanz.de (K.D.)
R E S E A R C H A R T I C L E S
www.sciencemag.org SCIENCE VOL 308 29 APRIL 2005 659

movement of the cytoplasmic against the peri-
plasmic surface of the c ring (fig. S1).
Vonck et al. (13) determined the position of
the periplasmic and the cytoplasmic interfaces
of the membrane surrounding the c ring, which
are approximately at Tyr80 and Ser55, respec-
tively (Fig. 2). On the inner surface of the c
ring, near the N- and C-termini, the map shows
an extended electron density that can be mod-
eled by an alkyl chain of nine C atoms per c
subunit. We propose that detergents are bound
here at a position corresponding to that of the
external leaflet of the membrane. The Phe5 resi-
dues at the periplasmic end of the inner surface
form a ring that possibly marks the position of
the glycerol backbone of phospholipids. Such
a positioning of lipids would correlate to the
‘‘central plug’’ feature seen in two-dimensional
crystals of c rings (23). Despite the lack of di-
rect evidence for a bilayer inside the c ring, its
existence can be inferred from the hydrophobic
surface that extends beyond the electron density
of the alkyl chains until Tyr34 (fig. S2B). How-
ever, the internal bilayer appears to be wider
and shifted toward the periplasmic surface
with respect to its external counterpart.
Structure of the sodium ion binding
site. The c ring was crystallized in buffer con-
taining 100 mM sodium acetate, which pro-
motes Naþ binding as shown by established
techniques (15). The Naþ ions are seen in the
map as 11 distinct densities at the bend of the
helices, close to the outer surface of the c ring.
Figure 2 shows a section of the c ring with two
C-terminal and one N-terminal helix forming a
Naþ-binding unit. The coordination sphere is
formed by side-chain oxygens of Gln32 (Oe1)
and Glu65 (Oe2) of one subunit and the hydrox-
yl oxygen of Ser66 and the backbone carbonyl
oxygen of Val63 of the neighboring subunit
(Figs. 2 and 3). The distance between the lig-
anding atoms and the ion is 2.37 T 0.14 ). The
liganding residues are in good accordance with
mutational studies, which recognized Gln32,
Glu65, and Ser66 as being essential for Naþ
binding (24), and their arrangement was
confirmed as a Naþ binding site using the
algorithm of Nayal and Di Cera (25). The crys-
tal structure shows the bound ion surrounded
by a network of hydrogen bonds: The Oe1 of
Glu65 accepts hydrogen bonds from the hy-
droxyl groups of Ser66 and Tyr70, and the Ne2
of Gln32 donates a hydrogen bond to Oe2 of
Glu65 (Fig. 3). These hydrogen bonds serve to
keep Glu65 deprotonated at physiological pH
(26) in order to allow Naþ binding and to lock
it into its ion-binding conformation. This ar-
rangement of residues and their hydrogen
bonds obviously serves to optimize the solva-
tion energy (27) of the Naþ ion and results in a
locked conformation of the ion binding site.
We propose that the present structure with
the side chain of Tyr70 facing outward and
stabilizing the side-chain conformation of
Glu65 represents the conformation outside of
the subunit a/c interface. Toward the peri-
plasmic surface from the binding site, the
structure forms a cavity (fig. S2B) to which
the side chain of Tyr70 could relocate to al-
low unloading and loading of the binding site
to and from subunit a. This relocation may
be part of an unlocking mechanism.
The Naþ binding signature is conserved in
all known Naþ-translocating ATP synthases
and appears in the c subunits of other an-
Fig. 1. Structure of the I. tartaricus c11 ring in
ribbon representation. Subunits are shown in
different colors. (A) View perpendicular to the
membrane from the cytoplasmic side. Two sub-
units are labeled. (B) Side view. The blue spheres represent the bound Naþ ions. Detergent
molecules inside the ring are shown with red and gray spheres for clarity. The membrane is
indicated as a gray shaded bar (width, 35 )). The images were created with PyMOL (36).
Fig. 2. Section of the c ring showing
the interface between the N-terminal
and two C-terminal helices with those
side chains discussed in the text. This
three-helix bundle represents a func-
tional unit responsible for Naþ bind-
ing and allowing access of the ion to
the binding site. The view is normal to
the external surface of the c ring with
the ring axis approximately vertical.
The color coding of the subunits is the
same as in Fig. 1A.
Fig. 3. Electron density map (red, Naþ omit
map at 3.0 s; blue, 2Fobs j Fcalc map at 1.4 s)
and residues of the Naþ binding site formed by
two c subunits, A and B. Naþ coordination and
selected hydrogen bonds are indicated with
dashed lines. Distances are given in ). The blue
sphere indicates the center of the bound Naþ
ion. The view is the same as in Fig. 2.
R E S E A R C H A R T I C L E S
29 APRIL 2005 VOL 308 SCIENCE www.sciencemag.org660