PERSPECTIVES
References and Notes
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
J. Fernandez, W. Ip, Planet. Space Sci. 44, 431 (1984).
R. Malhotra, Astron. J. 110, 420 (1995).
R. S. Gomes, Astron. J. 120, 2695 (2000).
The current mass of the Kuiper belt is only 0.01 to 0.1
Earth masses, but a few tens of Earth masses had to
exist in the region where the Kuiper belt bodies
formed in order for the growth time scale to be short
enough. Therefore, the current Kuiper belt apparently
contains only 0.1 to 1.0% of its primordial mass.
The inclination distribution of the nonresonant Kuiper
belt objects is bimodal (13). A first component, called
“cold population,” has inclinations smaller than about
4°. A second component, called “hot population,” has
a much broader inclination distribution, extending up
to 30° to 40°. The hot and cold populations seem to
have different physical properties (14, 15).
H. F. Levison, A. Morbidelli, Nature 426, 419 (2003).
R. S. Gomes, A. Morbidelli, H. F. Levison, Icarus 170,
492 (2004).
R. S. Gomes, Icarus 161, 404 (2003).
C. Trujillo, M. Brown, Astrophys. J. 554, 95 (2001).
Four plausible explanations have been proposed: (i) The
outer part of the disk was destroyed by the passage of
a star (16). (ii) It was photo-evaporated, owing to the
presence of massive stars in the neighborhood of the
Sun (17). (iii) The gas disk was extended, but planetesi-
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
mals beyond some threshold distance could not grow
because of the enhanced turbulence in the outer disk
(18). (iv) Distant dust particles and/or planetesimals
migrated to a smaller heliocentric distance during their
growth as a consequence of gas drag (19, 20).
W. K. Hartmann, G. Ryder, L. Dones, D. Grinspoon, in
Origin of the Earth and Moon, R. Canup, K. Righter,
Eds. (Univ. of Arizona Press, Tucson, AZ, 2000), pp.
493–512.
H. F. Levison, L. Dones, C. Chapman, A. Stern, M.
Duncan, K. Zahnle, Icarus 151, 286 (2001).
M. Brown, Astron. J. 121, 2804 (2001).
H. F. Levison, S. A. Stern, Astron. J. 121, 1730 (2001).
C. A. Trujillo, M. E. Brown, Astrophys. J. 566, 125
(2002).
S. J. Ida, J. Larwood, A. Burkert, Astrophys. J. 528, 351
(2000).
F. C. Adams, D. Hollenbach, D. Laughlin, U. Gurti,
Astrophys. J. 611, 360 (2004).
J. M. Stone, C. F. Gammie, S. A. Balbus, J. F. Hawley, in
Protostars and Planets IV, V. Mannings, A. P. Boss, S. S.
Russell, Eds. (Univ. of Arizona Press, Tucson, AZ, 1998),
pp. 589–612.
A. N. Youdin, F. H. Shu, Astrophys. J. 580, 494 (2002).
S. Weidenschilling, in Comets II, M. Festou, H. U. Keller,
H. Weaver, Eds. (Univ. of Arizona Press, Tucson, AZ, in
press).
Downloaded from www.sciencemag.org on August 30, 2007
Moon itself, during a cataclysm usually
called the “late heavy bombardment” (11).
What caused this bombardment, which occurred throughout the Solar System?
Previous studies (12) argued that a massive planetesimal disk in the outer Solar
System could have caused the late heavy
bombardment. In my opinion, there are not
many realistic alternatives to this explanation. However, according to our current understanding, after the migration of the giant
planets was over, the Solar System looked
essentially like the current one, with no
massive planetesimal populations left. It is
thus tempting to conjecture that the late
heavy bombardment was triggered by a late
start of planet migration. But why did migration start late, rather than soon after
planetary formation? The answer is not
known, yet.
Voltage Sensor Meets
Lipid Membrane
Roderick MacKinnon
lectrical impulses propagate rapidly
along the membranes of living cells.
The molecular components that make
this possible are proteins known as voltage-dependent ion channels. These channels open in response to changes in the
voltage across the cell membrane, and it is
precisely this voltage-dependent property
that allows them to propagate electrical impulses. Thirty years ago, Armstrong and
Bezanilla demonstrated that when voltagedependent ion channels experience a
change in the membrane voltage, tiny electrical charges known as gating charges
move relative to the membrane electric
field (1). This fundamental observation
suggested that a transmembrane voltage
change exerts an electric force on the gating charges, causing the pore within the
channel protein to open.
Now we know that voltage-dependent
potassium ion (K+) channels contain an ionselective pore domain with a gate, and a voltage sensor domain (segments S1 to S4) attached to the pore (see the figure). The gating
charges correspond to positively charged
arginine residues located on the otherwise
hydrophobic S4 segment. Thus, voltagesensing results from a repositioning of the
E
The author is in the Howard Hughes Medical Institute
and Laboratory of Molecular Neurobiology and
Biophysics, Rockefeller University, New York, NY
10021, USA. E-mail: [email protected]
1304
arginine residues within the membrane electric field that is associated with structural rearrangements of the voltage sensor, and these
structural rearrangements are linked to the
opening of the pore’s gate (2, 3).
The voltage sensor’s structure and the
process by which the gating charges are
repositioned have been subjects of intense
controversy. On the basis of electrophysiological studies, a number of structural models of the voltage-dependent K+ channel
have been proposed. These models share
the feature of an S4 helix that is isolated
from the lipid membrane by a protein wall
consisting of helices S1, S2, and S3 on the
channel’s lipid-facing perimeter (4–8).
They posit that a voltage change across the
membrane causes a translation or rotation
of the S4 helix, which would move the S4
helix and its positively-charged arginine
residues within an aqueous “gating pore.”
Recently, x-ray crystal structures (9), biotin-avidin accessibility studies (10), and
electron microscopy (11) of KvAP, a
prokaryotic voltage-dependent K+ channel,
have suggested a different model. In this
model (the paddle model) the voltage sensor is a highly mobile domain, and it is “inside-out” in the sense that helices S1, S2,
and S3 do not isolate S4 from the membrane; instead, S4 itself is located at the
protein-lipid interface. Specifically, S4 engages part of S3 to form a helix-turn-helix
“paddle” that could somehow move at the
19 NOVEMBER 2004
VOL 306
SCIENCE
Published by AAAS
protein-lipid interface. It is the location of
S4 that is at the center of the controversy. Is
S4 at the protein-lipid interface or is it
shielded from the lipid membrane by S1,
S2, and S3? The paddle model is based on
a collection of data—a full-length crystal
structure with obvious distortions of its
voltage sensor (12), a crystal structure of
the isolated voltage sensor (13), and accessibility data (10)—and thus is conceptual,
not atomic, and in many respects still needs
to be defined.
A recent report in Science by Perozo
and his co-workers (14) presents new data
on the structure of the KvAP voltage sensor. These authors studied the spin-label
side-chain accessibility and mobility of
KvAP K+ channels in lipid membranes using electron paramagnetic resonance
(EPR) spectroscopy. This is a particularly
informative technique for analyzing membrane proteins because it uses accessibility
parameters determined from the spectral
effects of lipid-soluble (O2) and water-soluble (NiEDDA) relaxing agents to distinguish between lipid-accessible and wateraccessible surfaces (15). A spin-label side
chain at a specific position on a protein can
thus be classified into one of three categories: buried beneath the protein surface,
on the surface exposed to aqueous solution, or on the surface exposed to lipid.
Furthermore, a side-chain mobility value
provides additional information; surface
positions tend to have a higher mobility
value than those buried inside the protein.
All voltage-dependent ion channels undergo conformational changes. Which conformation did Perozo and his colleagues
analyze? The KvAP voltage sensor is held
in a closed conformation when the voltage
is negative (for example, –100 mV) on the
inner membrane surface relative to the out-
www.sciencemag.org
CREDIT:
STRUCTURAL BIOLOGY
CREDIT: KATHARINE SUTLIFF/SCIENCE
PERSPECTIVES
17). The crystal structures of KvAP
provide an explanation for this: The
gating-sensitive arginines correWater
Paddle
spond precisely to those located on
Ion-conduction pore
S3
the voltage-sensor paddle, whereas
S4
the gating-silent arginines are found
off the paddle, on the so-called S4 to
*
**
S5 linker (9). Of the four paddle
*
arginines, the spin-label data suggest
Voltage
Protein
Lipid
sensor
that two (the third and fourth) are
buried within the protein and two
(the first and second) are exposed on
the surface. The first of these is exposed only to lipid, whereas the second is exposed to lipid and water
(see the asterisks in the figure).
+
A model of the KvAP K channel with a voltage-sensor
The new data from Perozo and his
at the protein-lipid interface. Accessibility and mobilicolleagues
are entirely consistent
+
ty data mapped onto a representation of the K channel
show the relation of the channel’s voltage sensor to the with the concepts proposed for the
lipid membrane, water surfaces, and pore (14). The figure open conformation of the paddle
depicts one possible explanation for the accessibility da- model (10). Many structural aspects
+
ta reported by Perozo and colleagues (14), with two ex- of the voltage-gated K channel are
posed and two buried positive charges in an apparently still uncertain. For instance, the new
open conformation of the voltage sensor. Four identical spin-label data by Perozo’s group do
subunits (brown) of the voltage-dependent K+ channel not address the question of how the
surround a central ion-conduction pore. Each subunit has voltage sensor moves. Fluorescence
a voltage sensor comprising α-helical segments S1 to S4. measurements of the Shaker K+ chanAccessibility measurements suggest that S1 and S2 are nel have led to the hypothesis of small
closest to the pore. In contrast, S3 and S4 reside at the (~2 Å) charge movements across a
protein-lipid interface, with one surface (yellow) exposed strong electric field that is highly foto lipid, another (red) exposed to protein, and with water cused by aqueous crevasses penetrat(blue) coating the “top.” The S4 helix contains four posi- ing the protein surface (3). In contively charged arginine residues (asterisks) that allow the trast, avidin accessibility experiments
channel to sense changes in membrane voltage.
on the KvAP channel suggest large
(at least 15 Å) movements of the voltpore. But the basic conclusion seems in- age-sensor paddle at the protein-lipid interescapable: S3 and S4 reside at the protein- face (10). There will no doubt be further dislipid interface, against the lipid membrane agreements over movements of the voltage
(see the figure).
sensor, and these will drive our understandIn the crystal structures of KvAP, the car- ing still further. What we need next are new
boxyl-terminal half of S3 and S4 form the structures, additional biochemical and funchelical-hairpin paddle of the voltage sensor tional analyses, accessibility data on the
(9). Although such an analysis was not pre- closed conformation of the channel, and a
sented in the paper, when the spin-label da- better chemical understanding of the prota are mapped onto the paddle a striking pat- tein-lipid interface. All of these are sure to
tern constrains the orientation of the paddle come our way in the near future.
in the membrane. As shown in the figure,
one face is exposed entirely to lipid (yelReferences
1. C. M. Armstrong, F. Bezanilla, J. Gen. Physiol. 63, 533
low), whereas the other is exposed to water
(1974).
near the hairpin turn (blue) and to a mixture
2. F. J. Sigworth, Q. Rev. Biophys. 27, 1 (1994).
of protein and lipid (red and yellow) further
3. F. Bezanilla, E. Perozo, Adv. Protein Chem. 63, 211
“down,” as though the paddle’s long axis
(2003).
4. R. Horn, J. Gen. Physiol. 120, 449 (2002).
were roughly perpendicular to the mem5. C. S. Gandhi, E. Y. Isacoff, J. Gen. Physiol. 120, 455
brane plane with one face toward the chan(2002).
nel and the other toward the membrane.
6. F. Bezanilla, J. Gen. Physiol. 120, 465 (2002).
Where are the arginine residues? The au7. H. P. Larsson, J. Gen. Physiol. 120, 475 (2002).
8. M. Laine et al. Neuron 39, 467 (2003).
thors state that four out of six are buried. Yet
9. Y. Jiang et al., Nature 423, 33 (2003).
the answer to this question with respect to
Y. Jiang et al., Nature 423, 42 (2003).
the positive charges involved in gating re- 10.
11. Q. Jiang et al., Nature 430, 806 (2004).
quires careful consideration. In the Shaker 12. Protein Data Bank (PDB) accession code 1ORQ.
voltage-dependent K+ channel isolated from 13. PDB accession code 1ORS.
fruit fly, the amino-terminal four S4 argi- 14. L. G. Cuello et al., Science 306, 491 (2004); published
online 15 October 2004 (10.1126/science.1101373).
nine residues contribute to the gating 15. W.
L. Hubbel et al., Curr. Opin. Struct. Biol. 8, 649 (1998).
charge, whereas the arginine residues near 16. S. A. Seoh et al., Neuron 16, 1159 (1996).
the carboxyl-terminal end of S4 do not (16, 17. S. K. Aggarwal, R. MacKinnon, Neuron 16, 1169 (1996).
www.sciencemag.org
+
Voltage-dependent K channel
SCIENCE
VOL 306
Published by AAAS
19 NOVEMBER 2004
Downloaded from www.sciencemag.org on August 30, 2007
side surface of the cell. The sensor moves
to its open conformation upon membrane
“depolarization” to 0 mV, causing the pore
to open. After a few seconds at 0 mV, the
pore becomes inactivated and ion conduction stops by an as yet unknown mechanism, although in all likelihood the voltage
sensor remains roughly in its open conformation. This is the condition under which
the EPR experiments were carried out by
Perozo’s group: membranes at 0 mV and
the voltage sensor presumably open.
What did they find? In a spin-label scan
of the voltage sensor (helical segments S1
to S4), the pattern of accessibility satisfies
expectations for a protein that spans the
membrane several times, that is, the waterexposed residues occur in the hydrophilic
loops between transmembrane segments.
This result is presented as being inconsistent with the paddle model, and the inconsistency is demonstrated through accessibility calculations made on an atomic version of a paddle model. This analysis is inappropriate because the paddle model, as
originally presented, contained uncertainties (particularly with respect to the positions of S1 and S2) that precluded calculations of atomic coordinates (10).
Nevertheless, the new data convey a
very telling message: Side-chain mobility
and lipid accessibility both increase as
measurements proceed from S1 to S4. As
the authors point out, this suggests that S1
and S2 are buried in a protein environment,
whereas S3 and S4 are more exposed to the
lipid membrane (see the figure).
Side-chain accessibility and mobility
data in the absence of distance constraints
are insufficient to deduce a complex protein
structure. But the authors propose a structural interpretation based on the assumption
that the crystal structure of the isolated voltage sensor (13) approximates a native conformation in the membrane lipid bilayer.
Given this assumption, they then ask which
orientation of the crystal structure of the
isolated sensor and pore best satisfy the
constraints imposed by the spin-label data?
Perozo and colleagues conclude that S1 and
S2 must represent the surface that lies
against the pore, away from the membrane,
thus accounting for their buried location.
They conclude that S3 and S4 helices reside
on the outer perimeter of the protein against
the membrane, thus accounting for their
high lipid exposure and high mobility. In reality, the voltage sensor in situ may adopt a
somewhat different structure than that of
the isolated voltage sensor. This is especially true at its covalent attachment site (S4 of
the voltage sensor to S5 of the pore) where
differences are implied by the spin-label data, and at the surface formed by S1 and S2,
which somehow forms an interface with the
1305