doi:10.1016/j.jmb.2007.03.063
J. Mol. Biol. (2007) 369, 413–418
Supramolecular Assembly of VDAC in Native
Mitochondrial Outer Membranes
Rui Pedro Gonçalves 1 , Nikolay Buzhynskyy 1 , Valérie Prima 2
James N. Sturgis 2 and Simon Scheuring 1 ⁎
1
Institut Curie,
UMR168-CNRS, 26 Rue
d'Ulm, 75248 Paris Cedex 5,
France
2
UPR-9027 LISM,
IBSM, CNRS,
31 Chemin Joseph Aiguier,
13402 Marseille Cedex 20,
France
The voltage-dependent anion channel (VDAC) is the most abundant protein
in the mitochondrial outer membrane (MOM). Due to its localization,
VDAC is involved in a wide range of processes, such as passage of ATP
out of mitochondria, and particularly plays a central role in apoptosis.
Importantly, the assembly of VDAC provides interaction with a wide range
of proteins, some implying oligomerization. However, many questions
remain as to the VDAC structure, its supramolecular assembly, packing
density, and oligomerization in the MOM is unknown. Here we report the
so far highest resolution view of VDAC and its native supramolecular
assembly. We have studied yeast MOM by high-resolution atomic force
microscopy (AFM) in physiological buffer and found VDAC in two distinct
types of membrane domains. We found regions where VDAC was packed at
high density (∼80%), rendering the membrane a voltage-dependent
molecular sieve. In other domains, VDAC has a low surface density
(∼20%) and the pore assembly ranges from single molecules to groups of up
to 20. We assume that these groups are mobile in the lipid bilayer and allow
association and dissociation with the large assemblies. VDAC has no
preferred oligomeric state and no long-range order was observed in densely
packed domains. High-resolution topographs show an eye-shaped VDAC
with 3.8 nm × 2.7 nm pore dimensions. Based on the observed VDAC
structure and the pair correlation function (PCF) analysis of the domain
architectures, we propose a simple model that could explain the phase
behavior of VDAC, and illustrates the sensitivity of the molecular
organization to conditions in the cell, and the possibility for modulation
of its assembly. The implication of VDAC in cytochrome c release from the
mitochondria during cell apoptosis has made it a target in cancer research.
© 2007 Elsevier Ltd. All rights reserved.
*Corresponding author
Keywords: atomic force microscopy; membrane protein; mitochondrion;
porin; voltage-dependent anion channel
Introduction
It is generally accepted that mitochondria evolved
from cell-enclosed symbiotic prokaryotes, therefore
mitochondria and prokaryotes share several characteristics.1 Both possess outer membranes with
densely packed porins that allow the passage of
relatively large molecules. Bacterial porins are
Abbreviations used: VDAC, voltage-dependent anion
channel; MOM, mitochondrial outer membrane; AFM,
atomic force microscopy; PCF, pair correlation function.
E-mail address of the corresponding author:
[email protected]
structurally well described,2 in contrast, many questions remain as the voltage-dependent anion channel (VDAC) structure, its supramolecular assembly,
packing density, and oligomerization in the mitochondrial outer membrane (MOM) are unknown.
VDAC is a 30 kDa protein found in the MOM of all
eukaryotes.3 Sequence, biochemical and computational analyses indicate a beta-barrel architecture
with a fold of 124 to 175 beta-strands and an Nterminal domain that is probably alpha-helical.5
Electron microscopy (EM) of 2D-crystals, readily
found in MOM preparations,6 or formed through
lipid withdrawal from MOM,7 or through VDAC
reconstitution,8 revealed the overall shape of the
molecule, a pore with an inner diameter of ∼3 nm,
0022-2836/$ - see front matter © 2007 Elsevier Ltd. All rights reserved.
414
from which a beta-barrel Cα-diameter of 3.6–3.8 nm
was estimated.
The sequence and function of VDAC are well
preserved among eukaryotes. It forms a single pore
with voltage dependence and anion selectivity.3
VDAC is the most abundant protein in the MOM
(>50% of total protein7) and is implicated in a large
number of processes beyond that of a simple anion
channel function: ATP transport, 9 superoxyde
anion release,10 and apoptosis11–13 among others.
It has been shown that interactions with cations
(Ca2+) control the gating properties of MOM and
VDAC.14 Furthermore, VDAC interacts with several mitochondrial and cytoplasmic proteins including kinases,15–17 cytochrome c,13 and actin,18 at
different binding sites. VDAC is believed to act
as an anchor point to proteins, which therefore have
an easier access to ATP produced by the mitochondria.16 In cell apoptosis,12 the function of VDAC
can be regulated by interacting with the members of
the Bcl-2 (B-cell leukemia/lymphoma 2) family,13,19
therefore VDAC has become a drug target for cancer
therapy.20,21 Recently, no less than 55 novel functional interaction partners of VDAC were found.22
The full extent and understanding of VDACs importance in the cells life and death is being extensively studied.23
Results and Discussion
Atomic force microscopy (AFM)24 has developed
into a powerful tool in membrane research.
Recently, topographs at ∼10 Å resolution of multicomponent native membranes from bacteria25 and
eukaryotes26 were acquired. Here, we used AFM to
study yeast MOM in physiological buffer, under
close-to-native conditions, reporting on the structure
and the supramolecular organization of VDAC in
native membranes. MOM were purified from yeast
essentially as previously described.7 The majority of
the membranes had vesicular shape (Figure 1(a))
with an area of ∼1 μm2 (Figure 1(b)). These
flattened vesicles contained featureless, presumably
lipid-only domains with an average thickness of
3.8 nm, and other domains densely packed with
pores whose diameters were consistent with those of
the mitochondrial porin VDAC. VDAC is the by far
most abundant protein in our preparation as
analyzed by Coomassie brilliant blue stained SDSPAGE and mass spectrometry (Figure 1(c)). In early
studies, poorly ordered 2D-crystals of VDAC were
readily found in MOM preparations.6
In medium resolution topographs, corrugated
pore-containing membrane areas were readily distinguished from the smoother, lipid-only regions,
allowing the estimation of pore packing densities.
Mixed domains contained pores at low (∼20%)
density, in other regions the pores were packed at
high (∼80%) density (Figure 2(a)). Often the pore
topography was accompanied by large protrusions
of ∼4 nm height (Figure 2(a)). Since VDAC is the
only outer membrane protein present at sufficient
Structure and Supramolecular Assembly of VDAC
Figure 1. MOM preparation morphology. (a) Overview: AFM topograph of MOM adsorbed on mica. (b)
MOM vesicles area histogram. (c) Coomassie brilliant blue
stained SDS-PAGE of yeast MOM preparation. Bands
labeled and marked with solid lines were identified by
liquid chromatography and mass spectrometry analysis,
and gel migration behavior. Bands marked with dotted
lines were not identified. VDAC is by far the most
abundant protein in the preparation.
quantity to account for this high density of pores, it is
safe to presume that the pores represent VDAC, and
so the protrusions are likely proteins that bind to
VDAC on the mica-facing side of the membrane.
Due to the large number of molecules that interact
with VDAC,18 it is difficult to ascertain which one it
might be. As it was suggested that VDAC's channel
behavior depends on its surface density,7 both the
high-density and the low-density domains are of
significant functional interest. In mixed domains,
VDAC monomers, dimers, trimers, hexamers, and
arrays of up to 20 molecules were found (Figure 2(b)
and (c)). These molecules are highly mobile and
difficult to image by AFM. High mobility of VDAC
groups in the mixed domains could trigger VDAC
association in the densely packed porous domains
and thus act as a regulator of VDAC activity. Such an
association–dissociation equilibrium is a simple way
of modulating channeling cooperativity7 and possible interactions that demand oligomerization.15,27
Little is known on VDACs structure. EM of poorly
ordered 2D-crystals allowed the calculation of lowresolution maps, showing an approximately circular
pore with dimension ∼3.7 nm (beta-barrel backbone
distance).7 Sequence analysis, analogy considerations with bacterial outer membrane porins and
circular dichroism studies all indicate that VDAC
would span the membrane as a beta-barrel (for a
review see Bay & Court5). The arrangement of the
Structure and Supramolecular Assembly of VDAC
415
not feature the N-terminal, possibly pore-filling domain, and that such a plug remained rather in the
inter-membrane-space half side of the channel.
Unlike in EM reconstructions,6 only one side of
the protein could be imaged. Electrostatic interactions with the negatively charged mica surface
may favor unidirectional adsorption of MOM. The
VDAC cytoplasmic surface is irregular with maximal protrusion of ∼1 nm over the bilayer surface,
corresponding to the beta-turns connecting the betastrands as was seen by EM and predicted by sequence analysis.
Pair correlation function (PCF) analysis revealed
interesting details about VDAC assembly in native
Figure 2. Low-density (∼20%) and high-density (∼80%)
VDAC domains in MOM. (a) Deflection image of a membrane patch containing low-density (L) and high-density
(H) domains. Protein corrugation (white outline) is easily
distinguishable from smooth lipid areas. Large protrusions
underlie VDAC in both low-density and high-density
domains (yellow outlines). (b) and (c) High-resolution
topographs of low-density domains. Outlines: monomeric
(m), dimeric (d), trimeric (t), hexameric (h), and arrays of
up to 20 VDAC (x) were found. The hexameric (h) VDAC
assembly resembles the unit cell arrangement found in
2D crystals.7
beta-strands, the loops and the N terminus, that may
form an alpha-helical central plug, is still a matter of
controversy. The proposed number of beta-strands
in the barrel ranges from 124 to 17.5 High-resolution
AFM imaging of the densely packed membranes
allowed us to obtain further structural information
(Figure 3(a)). VDAC is an eye-shaped pore with
dimensions of 3.8( ± 0.8) nm × 2.7( ± 0.6) nm (n = 98)
(Figure 3(b)). Some VDAC exhibited variability in
pore dimension that we assign to the protein's intrinsic flexibility. This flexibility may have hampered the growth of highly ordered crystals so far,
and could functionally be related to different conductance states. The AFM tip could enter up to 2 nm
inside some pores, indicating that the membranes
probably exposed their cytoplasmic face, that does
Figure 3. VDAC structure and supramolecular organization in densely packed domains. (a) High-resolution
AFM topograph of densely packed VDAC in a native
MOM. (b) High-resolution view of eye-shaped VDAC
(pore dimensions 3.8 nm × 2.7 nm; pore depth up to
2 nm). (c) PCF analysis of VDAC in densely packed domains. A peak at 53 Å represents the most frequent poreto-pore distance. A minor broad peak at ∼92 Å represents
rough hexagonal packing; no long-range order peak
was found (sketches represent corresponding pore-pore
assemblies).
416
MOM (Figure 3(c)): the most frequently found
assembly distance is 53 Å, corresponding to two
neighboring pores. This value is slightly larger than
43 Å to 50 Å,7 found in EM reconstructions. Importantly, the 53 Å distance represents the assembly
in the native membrane, whereas the shorter distances were obtained on 2D-crystals. A minor and
broad signal was found around 92 Å. This is √3
times the pore–pore neighbor distance, indicating
that VDAC is roughly in hexagonal packing. The
broadness of the 92 Å PCF peak is probably caused
by the elliptical shape of VDAC and a lack of
rotational alignment inducing distance dispersion
between molecules separated by one VDAC. No
long-range order peaks were found. In agreement,
calculated power spectra of these membrane regions did not reveal distinct diffraction peaks
corresponding to crystalline order. Lack of crystallinity is not contradictory with reports that predict
VDAC association with oligomeric kinases15–17 and
Bax,27 because the very high local density allows
this kind of binding. Indeed, only a non-crystalline
fluid assembly can account for assembly dynamics
that are demanded for the manifold interactions and
functional alterations of VDAC.
In an attempt to model the supramolecular
structure of VDAC with a minimum of parameters,
we have examined, using a Monte-Carlo approach,
potentials capable of driving the formation of
membrane domains resembling those that we have
observed. The presence of many small aggregates
indicates the absence of a macroscopic phase separation and a very dynamic system, strongly influenced by fluctuations and probably close to the
Structure and Supramolecular Assembly of VDAC
critical point. We used triangular potentials (hardcore diameter 53 Å) with long-range attractive forces
(diameter 106 Å) and potential well depths for
VDAC–VDAC interactions of 1.7 kBT to 2.0 kBT (kB
is the Boltzmann constant and T is the absolute
temperature). The resulting model (Figure 4(a)) fits
well many aspects of our analysis of the MOM.
Notably, the differentiation of high-density and lowdensity domains, a supramolecular assembly that
results in the same PCF peaks (Figure 4(b)) and the
presence of several small dynamic aggregates, are
reproduced. However, though this model is able to
reproduce some aspects of the VDAC assembly in
the MOM, it is very sensitive to the parameters used,
in particular the interaction potential depth. Thus if
this model is realistic it predicts also a strong sensitivity of VDAC organization to parameters that
might modulate protein–protein interaction strength
such as pH, membrane potential, ion concentration
and interactions with other proteins. It is particularly
noteworthy that in this case, as in our previous
studies of bacterial photosynthetic membranes,25 a
strong lateral organization and differentiation of
distinct membrane regions appear to be driven by
relatively weak (a few times kBT) long-range forces.
The origin of such long-range forces has not yet been
investigated, but membrane elastic and solvation
forces are probably of particular importance.
AFM imaging of MOM has allowed us to study
the supramolecular organization of VDAC and to
gain further insight into the protein structure. We
show how the mitochondrial porin can associate
in mobile groups of highly variable numbers, probably in a function-related manner that needs further
Figure 4. Modeling of the supramolecular assembly of VDAC in native MOM. (a) VDAC distribution model in MOM.
The separation of VDAC in high-density and low-density domains and the existence of small aggregates ranging from
single molecules up to about 20, are reproduced (compare to Figure 1). (b) PCF analysis of the modeled VDAC assembly.
As in the experimental data, a peak at 53 Å represents the most frequent intermolecular distance, and a minor peak at
∼92 Å represents rough hexagonal packing in high-density domains (compare to Figure 2(c)).
417
Structure and Supramolecular Assembly of VDAC
elaboration. VDAC also forms densely packed
patches, which may facilitate docking of other oligomeric proteins.
Materials and Methods
MOM purification
Mitochondria were purified from yeast (Saccharomyces
cerevisiae strain W303) grown overnight on YP-glycerol
(10 g/l yeast extract, 10 g/l peptone, 20 g/l glycerol)
media at 30 °C. Cells were harvested and washed
successively with water, Tris-SO4 buffer (100 mM Tris
H2SO4 (pH 9.4), 10 mM DTT) and sorbitol-buffer (1.2 M
sorbitol, 20 mM KPO4 (pH 7.4)). The cell wall was then
digested with 20,000 units of zymolase (Sigma) per gram
fresh weight of cells for 2 h at 30 °C. Spheroplasts were
collected and washed twice with fresh ice-cold sorbitolbuffer before being resuspended in breaking buffer (0.6 M
sorbitol, 20 mM Hepes-KOH (pH 7.4), 1 mM PMSF) and
lysed by 15 strokes with a Dounce homogenizer. Cells
debris were sedimented by centrifugation at 1500g for
5 min and mitochondria collected by centrifugation at
12,100g for 10 min and then washed once in the same
buffer. Mitochondria were quick frozen in liquid nitrogen
and stored at −80 °C. Outer membranes were prepared
essentially as described.7 Briefly, the mitochondrial pellet
was resuspended in ice-cold lysis medium (0.25 mM
EDTA, 0.25 mM EGTA, adjusted to pH 7.0 with NaOH),
undispersed material was removed by centrifugation at
1000g for 5 min, the solution was homogenized by three
strokes of a Dounce homogenizer and stirred for 25 min on
ice. Solid manitol was then added to a final concentration
of 5% and the resulting solution layered onto a 0.55 M–
0.9 M sucrose step gradient. After centrifugation at
60,000g for 90 min in a Beckman SW28 rotor the outer
membranes were recovered at the 0.55 M–0.9 M sucrose
interface, diluted five times with 1 mM Tris-HCl (pH 7.4)
and collected by centrifugation at 60,000g for 45 min and
washed with the same buffer. SDS-PAGE and mass
spectrometry were used to characterize the MOM protein
content.
Atomic force microscopy
The AFM24 was operated in contact mode at ambient
temperature and pressure. Imaging was performed with a
commercial Nanoscope-E AFM (Veeco, Santa Barbara,
CA, USA) equipped with a vertical engagement 160 μm
J-scanner and oxide-sharpened Si3N4 cantilevers (length
100 μm; k = 0.2 N/m; determined by the thermal calibration method on a Picoforce AFM). For imaging minimal
loading forces of ∼100 pN were applied, at scan
frequencies of 4–7 Hz using optimized feedback parameters. 3 μl of membrane solution (0.1 mg/ml of protein
concentration) were injected into a 20 μl adsorption buffer
drop (10 mM Tris-HCl (pH 7.2), 150 mM KCl, 25 mM
MgCl2) on freshly cleaved mica. Subsequently, after ∼2 h
of adsorption, the sample was rinsed with recording
buffer (10 mM Tris-HCl (pH 7.2), 150 mM KCl).
Data analysis
All image treatment and analysis of AFM topographs
were performed using self written routines for the ImageJ
image processing package28 and IGOR PRO (Wavemetrics, Lake Oswego, OR 97035 USA). Surface density
calculations were made using the lipid surface height as
threshold, and comparison of the surface occupied by
VDAC with the protein free surface. Pore dimensions
were measured by examining height profiles of VDAC
topographs, setting the lipid bilayer height as pore
delimiting threshold.
For pair correlation function analysis,29 pore center x
and y coordinates were defined using cross-correlation
routines.28 From this, separation vectors r, between pores
were calculated. Hence, the probability p(r) of finding a
complex (B) in a distance interval from r1 to r2 from a
given complex (A) can be normalized, taking into account
the surface area covered within the interval and the total
number of complexes within the membrane. If the
distribution is random, p(r) is always close to 1.29 For
each complex, the distance to the closest membrane border
was set as the maximum search radius, rmax.
Modeling
A Monte-Carlo approach was used to study the VDAC
assembly in MOM. A 2D random mixture of discs at a
density corresponding to the observed was used as
starting condition. The modeling was performed over
108 Monte-Carlo steps in the NVT ensemble. Step sizes for
displacements were updated regularly to maintain
approximate 50% success rates. The interaction potentials
used incorporated the observed hard-core diameters of
53 Å for VDAC, and an attractive long-range triangular
potential well that falls to zero at a distance of 106 Å. This
choice of potential distance is based on the observed lack
of long-distance order in dense regions, and our knowledge of the phase diagram. The well depth was adjusted to
approximate the observed PCF, paying particular attention to the relative heights of the successive peaks, with a
well depth of 1.7 kBT to 2.0 kBT for VDAC–VDAC
interactions providing kinetic trapping of the assembly
and disassembly of associations into aggregates of a few
molecules up to densely packed domains, in agreement
with our measurements. With weaker attractive forces
(e.g. 1.4 kBT) the dense regions were much less ordered
and the second peak in the PCF was considerably too low,
while stronger attractive forces (e.g. 2.2 kBT) resulted in
multiple strong long-range oscillations in the PCF.
Acknowledgements
This study was supported by the INSERM and
INSERM Avenir (to S.S.), a Ministère de l'Education
Nationale scholarship (to R.P.G.) and the ANR-06NANO-023-01 grant (to S.S.).
References
1. Wallin, I. E. (1922). A note on the morphology of
bacteria symbiotic in the tissues of higher organisms.
J. Bacteriol. 7, 471–474.
2. Cowan, S. W., Schirmer, T., Rummel, G., Steiert, M.,
Ghosh, R., Pauptit, R. A. et al. (1992). Crystal
structures explain functional properties of two E. coli
porins. Nature, 358, 727–733.
418
Structure and Supramolecular Assembly of VDAC
3. Colombini, M. (1980). Structure and mode of action of
a voltage-dependent anion-selective channel (VDAC)
located in the outer mitochondrial membrane. Ann.
N.Y. Acad. Sci. 341, 552–563.
4. Blachly-Dyson, E., Peng, S., Colombini, M. & Forte, M.
(1990). Selectivity changes in site-directed mutants of
the VDAC ion channel: structural implications.
Science, 247, 1233–1236.
5. Bay, D. C. & Court, D. A. (2002). Origami in the outer
membrane: the transmembrane arrangement of mitochondrial porins. Biochem. Cell Biol. 80, 551–562.
6. Mannella, C. A. (1982). Structure of the outer
mitochondrial membrane: ordered arrays of porelike
subunits in outer-membrane fractions from Neurospora crassa mitochondria. J. Cell Biol. 94, 680–687.
7. Mannella, C. A. (1998). Conformational changes in the
mitochondrial channel protein, VDAC, and their
functional implications. J. Struct. Biol. 121, 207–218.
8. Dolder, M., Zeth, K., Tittmann, P., Gross, H., Welte, W.
& Wallimann, T. (1999). Crystallization of the human,
mitochondrial voltage-dependent anion-selective
channel in the presence of phospholipids. J. Struct.
Biol. 127, 64–71.
9. Rostovtseva, T. & Colombini, M. (1996). ATP flux is
controlled by a voltage-gated channel from the
mitochondrial outer membrane. J. Biol. Chem. 271,
28006–28008.
10. Han, D., Antunes, F., Canali, R., Rettori, D. &
Cadenas, E. (2003). Voltage-dependent anion channels
control the release of the superoxide anion from
mitochondria to cytosol. J. Biol. Chem. 278, 5557–5563.
11. Tsujimoto, Y. & Shimizu, S. (2000). VDAC regulation
by the Bcl-2 family of proteins. Cell Death Differ. 7,
1174–1181.
12. Shimizu, S., Matsuoka, Y., Shinohara, Y., Yoneda, Y. &
Tsujimoto, Y. (2001). Essential role of voltage-dependent anion channel in various forms of apoptosis in
mammalian cells. J. Cell Biol. 152, 237–250.
13. Shimizu, S., Narita, M. & Tsujimoto, Y. (1999). Bcl-2
family proteins regulate the release of apoptogenic
cytochrome c by the mitochondrial channel VDAC.
Nature, 399, 483–487.
14. Bathori, G., Csordas, G., Garcia-Perez, C., Davies, E.
& Hajnoczky, G. (2006). Ca2+-dependent control of
the permeability properties of the mitochondrial
outer membrane and voltage-dependent anionselective channel (VDAC). J. Biol. Chem. 281,
17347–17358.
15. Zalk, R., Israelson, A., Garty, E. S., Azoulay-Zohar,
H. & Shoshan-Barmatz, V. (2005). Oligomeric states
of the voltage-dependent anion channel and cytochrome c release from mitochondria. Biochem. J. 386,
73–83.
16. Mathupala, S. P., Ko, Y. H. & Pedersen, P. L. (2006).
Hexokinase II: cancer's double-edged sword acting as
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
both facilitator and gatekeeper of malignancy when
bound to mitochondria. Oncogene, 25, 4777–4786.
Schlattner, U., Dolder, M., Wallimann, T. & TokarskaSchlattner, M. (2001). Mitochondrial creatine kinase
and mitochondrial outer membrane porin show a
direct interaction that is modulated by calcium. J. Biol.
Chem. 276, 48027–48030.
Roman, I., Figys, J., Steurs, G. & Zizi, M. (2006). Direct measurement of VDAC-actin interaction by surface plasmon resonance. Biochim. Biophys. Acta, 1758,
479–486.
Vander Heiden, M. G., Li, X. X., Gottleib, E., Hill, R. B.,
Thompson, C. B. & Colombini, M. (2001). Bcl-xL promotes the open configuration of the voltage-dependent anion channel and metabolite passage through
the outer mitochondrial membrane. J. Biol. Chem. 276,
19414–19419.
Lai, J. C., Tan, W., Benimetskaya, L., Miller, P.,
Colombini, M. & Stein, C. A. (2006). A pharmacologic
target of G3139 in melanoma cells may be the
mitochondrial VDAC. Proc. Natl Acad. Sci. USA, 103,
7494–7499.
Simamura, E., Hirai, K. I., Shimada, H., Koyama, J.,
Niwa, Y. & Shimizu, S. (2006). Furanonaphthoquinones cause apoptosis of cancer cells by inducing the
production of reactive oxygen species by the mitochondrial voltage-dependent anion channel. Cancer
Biol. Ther. 5.
Roman, I., Figys, J., Steurs, G. & Zizi, M. (2006).
Hunting interactomes of a membrane protein. Mol.
Cell. Proteomics, 5, 1667–1680.
Wang, X. (2001). The expanding role of mitochondria
in apoptosis. Genes Dev. 15, 2922–2933.
Binnig, G., Quate, C. F. & Gerber, C. (1986). Atomic
force microscope. Phys. Rev. Letters, 56, 930–933.
Scheuring, S. & Sturgis, J. N. (2005). Chromatic
adaptation of photosynthetic membranes. Science,
309, 484–487.
Buzhynskyy, N., Hite, R. K., Walz, T. & Scheuring, S.
(2007). The supramolecular architecture of junction
microdomains in native lens membranes. EMBO
Rep. 8, 51–55.
Mikhailov, V., Mikhailova, M., Pulkrabek, D. J., Dong,
Z., Venkatachalam, M. A. & Saikumar, P. (2001). Bcl-2
prevents bax oligomerization in the mitochondrial
outer membrane. J. Biol. Chem. 276, 18361–18374.
Scheuring, S., Boudier, T. & Sturgis, J. N. (2007).
From high-resolution AFM topographs to atomic
models of supramolecular membrane protein assemblies. J. Struct. Biol. In the press. doi:10.1016/j.
jsb.2007.01.021.
Baksh, M. M., Jaros, M. & Groves, J. T. (2004).
Detection of molecular interactions at membrane
surfaces through colloid phase transitions. Nature,
427, 139–141.
Edited by J. Bowie
(Received 5 February 2007; received in revised form 26 March 2007; accepted 26 March 2007)
Available online 30 March 2007