Like-charge attraction between polyelectrolytes

induced by counterion charge density waves

Thomas E. Angelini†, Hongjun Liang†, Willy Wriggers‡, and Gerard C. L. Wong†§
†Departments of Materials Science and Engineering, Physics, and Bioengineering, University of Illinois at Urbana–Champaign, Urbana, IL 61801; and
‡Laboratories for Biocomputing and Imaging, School of Health Information Sciences, University of Texas Health Science Center, 7000 Fannin Street,
Suite 600, Houston, TX 77030

Communicated by David Pines, Los Alamos National Laboratory, Los Alamos, NM, June 3, 2003 (received for review April 9, 2003)

Electrostatics in aqueous media is commonly understood in terms chosen as a model system because it is a rigid polyelectrolyte
of screened Coulomb interactions, where like-charged objects, (persistence length ␰A ⬇10 ␮m) with a large diameter (DA ⬇
such as polyelectrolytes, always repel. These intuitive expectations 75 Å) (12, 13), so that correlated counterion ‘‘lattices’’ can have
are based on mean field theories, such as the Poisson–Boltzmann the potential to form large domains, and thereby generate sharp,
formalism, which are routinely used in colloid science and compu- detectable peaks. Because the number of atoms that comprise
tational biology [Israelachvili, J. (1992) Intermolecular and Surface the charged biopolymer is large compared with the number of
Forces (Academic, London), 2nd ed.]. Like-charge attractions, how- counterions, high Z counterions (Ba2⫹) have been used to
ever, have been observed in a variety of systems [Gelbart, W. M., improve the Z-contrast between counterion scattering and pro-
Bruinsma, R. F., Pincus, P. A. & Parsegian, V. A. (2000) Phys. Today tein scattering. Finally, to further highlight the diffraction fea-
53, 38 – 44]. Intense theoretical scrutiny over the last 30 years tures that are generated by the counterions, we have compared
suggests that counterions play a central role, but no consensus the SAXS patterns from F-actin that has been condensed
exists for the precise mechanism. We have directly observed the nonelectrostatically (no added multivalent ions, prepared by
organization of multivalent ions on cytoskeletal filamentous actin progressive ultracentrifugation and water removal) with F-actin
(a well defined biological polyelectrolyte) by using synchrotron that has been condensed electrostatically by using multivalent
x-ray diffraction and discovered an unanticipated symmetry-break- ions. Unlike nonelectrostatically condensed samples, multivalent
ing collective counterion mechanism for generating attractions. ion condensed biopolymer samples typically exhibit poor align-
Surprisingly, the counterions do not form a lattice that simply ment, because of the existence of strong attractions and the
follows actin’s helical symmetry; rather, the counterions organize resultant precipitation of the biopolymer. We mitigated this
into ‘‘frozen’’ ripples parallel to the actin filaments and form 1D problem by using a small (300 ⫻ 300 ␮m2) synchrotron x-ray
charge density waves. Moreover, this 1D counterion charge density beam on a sheared sample.
wave couples to twist distortions of the oppositely charged actin Monomeric actin (G-actin) (molecular weight, 43,000) was
filaments. This general cooperative molecular mechanism is anal- prepared from a lyophilized powder of rabbit skeletal muscle
ogous to the formation of polarons in ionic solids and mediates purchased from Cytoskeleton (Denver). The nonpolymerizing
attractions by facilitating a ‘‘zipper-like’’ charge alignment be- G-actin solution contained a 5 mM Tris buffer at pH 8.0, with
tween the counterions and the polyelectrolyte charge distribution. 0.2 mM CaCl2, 0.5 mM ATP, 0.2 mM DTT, and 0.01% NaN3.
We believe these results can fundamentally impinge on our gen- G-actin (2 mg兾ml) was polymerized into F-actin (linear charge
eral understanding of electrostatics in aqueous media and are density ␭A ⬇ ⫺1e兾2.5 Å at pH 7) with the addition of salt (100
relevant to a wide range of colloidal and biomedical processes. mM KCl). Human plasma gelsolin (Cytoskeleton) was used to
regulate the F-actin length. The filaments were then treated with
phalloidin (molecular weight 789.2) to prevent actin depolymer-
A ll proposed theoretical explanations for like-charge attrac-
tion (1–10) introduce some form of positional counterion
correlations within a condensed layer of oppositely charged
ization. F-actin gels were ultracentrifuged at 100,000 ⫻ g to pellet
the filaments. After the removal of the supernatant buffer
solution, the F-actin was resuspended to a final concentration of
counterions (11). Dynamic ‘‘van der Waals’’-like correlations of
10 mg兾ml by using Millipore H2O (18.2 M⍀). Electrostatically
long-wavelength ion fluctuations have been suggested (2–5).
condensed samples were prepared by mixing with MgCl2 or
Static mechanisms consisting of ion correlations along the axis
BaCl2 salt solutions. All samples were sealed in 1.5-mm-diameter
of the polyelectrolyte rods in the form of a Wigner lattice have
quartz capillaries. SAXS experiments were performed at Beam-
also been considered (7–10). All extant theories, however, use
line 4-2 of the Stanford Synchrotron Radiation Laboratory. The
highly idealized models for the polyelectrolytes, such as infinitely
x-ray energy was monochromatized to 8.98 KeV by using a
thin lines or ‘‘smooth’’ rods. The present experimental work
double-bounce Si(111) crystal.
examines the extent to which these predictions survive in a real
polyelectrolyte system and provides a point of contact between Results and Discussion
the two extreme viewpoints above. The general form of the The basic supramolecular structure of the biopolymer compo-
counterion mechanism we observe here, which involves a cou- nent of the composite counterion-biopolymer system can be seen
pling between counterion correlation and polyelectrolyte defor- in the progressive formation of F-actin bundles with increasing
mation (Fig. 1), is applicable to a wide range of colloidal and multivalent ion concentrations. SAXS measurements of F-actin
biomedical processes that depend on controlling interactions condensed with Ba2⫹ (Fig. 2A) indicate an actin-actin close-
between macroions dispersed in fluid suspensions. packing peak characteristic of the bundled phase appears at
Materials and Methods q ⫽ 0.089 Å⫺1. These F-actin bundles can also be imaged in real
space by using laser-scanning confocal fluorescence microscopy
Diffraction features generated by counterion correlations are (Fig. 2B). The F-actin filaments were labeled with phalloidinated
expected to be weak compared with those generated by biopoly-
mer organization. To detect positional correlations in counte-
rion distributions on the surface of a charged polymer with small Abbreviations: F-actin, filamentous actin; CDW, charge density wave; sd-1, subdomain 1;
angle x-ray scattering (SAXS), a number of experimental strat- LRO, long range order; SAXS, small angle x-ray scattering.
egies have been used. Filamentous actin (F-actin) has been §To whom correspondence should be addressed. E-mail: gclwong@uiuc.edu.

8634 – 8637 兩 PNAS 兩 July 22, 2003 兩 vol. 100 兩 no. 15 www.pnas.org兾cgi兾doi兾10.1073兾pnas.1533355100
 aligned along the qz axis and are powder-averaged in the plane
perpendicular to the qz axis. For the analysis of the 2D SAXS
patterns (Fig. 3 A and D), cuts along qr and qz are produced by
integrating the 2D SAXS image over 40° wedges in ␹, centered
about the qr and qz axes, where qr ⬅ qcos␹, qz ⬅ qsin␹, q ⫽
(4␲兾␭)sin␪, and 2␪ is the Bragg angle. The three equatorial peaks
(a–c) in Fig. 3B correspond to the first-order, 公3-order, and
second-order peaks expected from local hexagonal coordination.
The two strong, partially orientationally averaged peaks (d and
e) along the meridional direction (Fig. 3C) are the well known
sixth- and seventh-layer lines characteristic of 13兾6 symmetry (13
monomers in 6 full turns) (14), in agreement with the known
structure of F-actin. To account for the sample mosaic and the
effects of integration on layer lines, we partially powder-
averaged (␲兾2 ⱖ ␹ ⱖ ␲兾4) theoretical 2D diffraction patterns.
The resulting modeled peak positions have been optimized by
varying the monomer spacing in 0.1-Å increments. These peak
positions and combined full width at half-maximum are well
described (to within 1%) by a Fourier transform of the standard
Fig. 1. Schematic representations of uncondensed and condensed F-actin. four-sphere model (15) for the actin filament, using the Holmes
(A) At low multivalent ion concentrations, two F-actin filaments maintain coordinates (14), combined with Gaussian layer line profiles (16).
their native 13兾6 symmetry and are unbound. (B) At high multivalent ion The diffraction pattern for multivalent ion condensed F-actin
concentrations, the ions collectively form a CDW and bundle F-actin filaments. bundles differs dramatically from that of the nonelectrostatically
Moreover, the CDW forms a coupled mode with torsional distortions of the condensed F-actin. Fig. 3D is a 2D SAXS pattern of a partially
F-actin and has overtwisted it by 3.8° per monomer to a new 36兾17 symmetry. aligned sample of F-actin at 7.5 mg兾ml condensed with 60 mM
The representations of F-actin are low-resolution density maps generated by
Ba2⫹. The filaments are oriented parallel to qz on average as in
using SITUS software (http:兾兾situs.biomachina.org).
the nonelectrostatically condensed sample above. The most
salient difference between the two diffraction patterns (between
Alexa Fluor 488 (Molecular Probes). A Leica SP-2 confocal Fig. 3 A and D and between Fig. 3 C and F) is the appearance
microscopy system (Beckman Institute, Urbana, IL) was used to of a sharp diffraction feature at qz ⫽ 0.105 Å⫺1 in the electro-
statically condensed case (Fig. 3F, peak i). This finding is

APPLIED PHYSICAL
image the dilute F-actin solutions in sealed sample cells at
surprising because the sample alignment is weaker for the
different ionic strengths. The observed thickness of bundles in

SCIENCES
electrostatically condensed actin compared with the nonelectro-
the image indicates that each bundle is comprised of thousands
statically condensed actin, and weaker alignment can only
of filaments. broaden and weaken existing peaks, but never produce new ones.
A direct comparison between the 2D SAXS patterns of F-actin It is in fact not possible to generate this new peak by using the
condensed osmotically without using divalent ions and of F-actin F-actin helix. This peak, which only occurs for multivalent ion
condensed electrostatically with divalent ions is made in Fig. 3. condensed samples and is aligned along the qz-axis, corresponds
Fig. 3A is a 2D SAXS pattern of partially aligned F-actin at 90 to ‘‘ripples’’ of multivalent ion density along the filament axis,
mg兾ml in the absence of divalent ions. The F-actin rods are analogous to a classical 1D charge density wave (CDW). Com-
pared to any of the F-actin layer line peaks in this region of
reciprocal space, this peak is symmetrical, significantly sharper,
and more intense. This peak is consistent with the diffraction
from a new 1D periodic charge distribution with a finite in-plane
width and azimuthal symmetry. Its diffraction signal is propor-
tional to the modulus squared of its Fourier transform, which is
the product of a periodic arrangement of Bragg sheets multiplied
by 0th-order Bessel functions centered on the qz axis, which
suppress the off-axis Bragg sheet intensity. Unlike the higher-
order Bessel function peaks from the F-actin helix, which are
shifted and broadened along qz by the large mosaic distribution
of bundle orientations within the sample volume, this peak
remains sharp along qz even after such orientational averaging.
Because we observe a single peak, the counterion CDW is
dominated by one Fourier component and is approximately
described by a simple sinusoidal density variation with a spatial
period of 59.8 Å, which is comparable to the spacing between
monomer ridges on the surface of one side of an F-actin filament
(approximately two times the monomer spacing). Because of the
weak diffraction signal of the CDW, we cannot rule out the
Fig. 2. (A) SAXS profiles of F-actin (⬇2 ␮m long) condensed by Ba2⫹ ions. At possibility of higher harmonics. Our results, however, suggest
0 mM Ba2⫹ concentration, F-actin is in an uncondensed isotropic phase. At that the dominant Fourier component of the CDW is the one at
high Ba2⫹ concentrations (36 and 108 mM), F-actin condenses into close
qz ⫽ 0.105 Å⫺1, which implies that the counterion distribution can
packed bundles with a correlation peak at q ⫽ 0.089 Å⫺1, indicating the
existence of an attraction. A weak higher-order reflection can also be ob-
be approximated by a ‘‘frozen’’ 1D sinusoidal ripple along the
served at q ⫽ 0.136 Å⫺1, which deviates slightly from the position expected polyelectrolyte.
from an exact hexagonal lattice. (B) A confocal image of a dilute (0.03 mg兾ml) This simple picture, however, is incomplete. For example, why
solution of bundles of F-actin (⬇10 ␮m) at 72 mM global Ba2⫹ concentration. do the counterions follow a new 1D symmetry rather than the full
(Scale bar, 8 ␮m.) helical symmetry of the charge distribution on F-actin? A close

Angelini et al. PNAS 兩 July 22, 2003 兩 vol. 100 兩 no. 15 兩 8635
Fig. 3. (A) A 2D SAXS pattern of nonelectrostatically condensed F-actin. The procedure for generating Integrated cuts along qr (B) and along qz (C) are described
in the text. Peak positions for all observed layer lines (6th, 7th, and 13th) agree with the native 13兾6 actin model to within ⬍1%. (D) 2D SAXS of Ba2⫹ condensed
F-actin. Cuts along qr (E) and qz (F) are performed the same way. The data indicate the formation of a counterion 1D CDW and an induced 3.8° overtwist of F-actin
(see text). The peak positions are 0.035 Å⫺1 (a), 0.062 Å⫺1 (b), 0.07 Å⫺1 (c), 0.111 Å⫺1 (d), 0.123 Å⫺1 (e), 0.225 Å⫺1 (f), 0.089 Å⫺1 (g), 0.136 Å⫺1 (h), 0.105 Å⫺1 (i),
0.115 Å⫺1 (j), 0.120 Å⫺1 (k), 0.136 Å⫺1 (l), and 0.227 Å⫺1 (m). (Inset) A magnification of the CDW peak (i) and neighboring layer line scattering (j and k).

examination of the diffraction pattern from the multivalent with the standard four-sphere model in the 13兾6 helix and
ion-condensed F-actin reveals that the 1D CDW has enforced a monomer spacing of 28.7 Å (the value extracted from the
change in the native F-actin symmetry and overtwisted the measurement of actin in Fig. 3A) and twisting the filament over
F-actin helix by 3.8° per monomer. This can be seen by the direct a range of ⫾10° per monomer in increments of 0.05°. Long repeat
comparison of Fig. 3D with the nonelectrostatically condensed helical symmetries up to the 108兾51 range have been checked for
actin in Fig. 3A. The equatorial peak positions (g and h) deviate the appearance of new intense layer lines, and none were found.
slightly from those expected from an exact hexagonal lattice and The measured peaks at q ⫽ 0.115 and 0.120 Å⫺1 agree well with
indicate that the bundled filaments are tightly packed (Fig. 3E). the l ⫽ 17 and 19 layer lines of the overtwisted 36兾17 symmetry
Along the meridional direction, the Bessel function peaks of the (calculated peaks at q ⫽ 0.113 and 0.120 Å⫺1 after adding mosaic
native 13兾6 layer lines of the nonelectrostatically condensed smearing). Even the higher-order feature at q ⫽ 0.227 Å⫺1 is
sample (Fig. 3C) have been replaced by a cluster of weaker peaks reproduced.
(j–l) in the multivalent ion condensed sample (Fig. 3F). The peak Why does the composite polyelectrolyte-counterion system
at 0.136 Å⫺1 (l) is clearly caused by the mosaic smearing of the choose the overtwisted 36兾17 symmetry? The angle between
strong transverse inter-rod correlation at the same q-value along adjacent monomers on a 13兾6 helix is given by ⫺6*2␲兾13, or
the equatorial direction, and the intensity agrees well with ⫺166.2°. By contrast, the angle per monomer on a 36兾17 helix
calculations accounting for the mosaic distribution. The remain- is ⫺170°, which indicates an additional twist of ⫺3.8° per
ing peaks can be completely explained by overtwisting F-actin to monomer. An examination of the local environment around a 1D
a new 36兾17 symmetry. Experimentally, F-actin is known to CDW reveals why the 36兾17 symmetry is favored. Consider a
exhibit a range of twists states (13, 17, 18), an observation that simple columnar hexagonal array of F-actin filaments decorated
is consistent with its low torsional rigidity, as measured with by parallel 1D CDWs. The most elementary unit of this structure
single-molecule techniques (19). We consider changes in the is a three-filament bundle, which defines a ‘‘channel’’ of trian-
F-actin helical twist away from its native symmetry by starting gular cross section, threaded by a 1D CDW of ions. Fig. 4 depicts

Fig. 4. For the 36兾17 and 13兾6 symmetries, a schematic representation of a three-filament bundle with the central triangular tunnel highlighted (Upper) and
the same bundle in its unwrapped form, showing the interior charge distribution of the tunnel (Lower). The red-colored regions correspond to the highly charged
sd-1s of the actin monomers facing the interior of the bundle. Double lines mark the location of the electrostatic兾steric defects where sd-1s from adjacent
filaments are in close proximity. Overtwisting the 13兾6 helix to a 36兾17 helix dramatically reduces the density of these defects and generates long regions of evenly
spaced charged patches to match with the periodic counterion CDW.

8636 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.1533355100 Angelini et al.

the negative charge distribution within such a three-filament in the diffraction signal to detect inter-CDW coupling, by
bundle. The red-colored regions on each filament correspond to discriminating between true long-range order (LRO), which is
subdomain-1s (sd-1s) of the actin monomers that face the inside expected if the CDWs in different channels are coupled, and
of the channel. These subdomains are the most highly negatively quasi-LRO (QLRO), which is expected from a 1D system (20).
charged and the most protrusive regions on the surface of In fact, the peaks are expected to be sharp for both LRO and
F-actin. To lower the electrostatic energy, these domains should QLRO, and the differences are detected usually in the asymp-
ideally be positioned adjacent to the maxima of the oppositely totic behavior of the correlation peaks (ref. 21 and references
charged 1D CDW. However, because F-actin is a helix, the highly therein).
charged sd-1 will process in and out of this central channel and
will not always be in register with the CDW. If we examine the Conclusions
‘‘unwrapped’’ three-filament bundle and follow sd-1 from mono- We have experimentally examined the mechanism for like-
mer to monomer and filament to filament along the bundle, we charge attraction in cytoskeletal F-actin and found evidence for
find that no matter how the initial relative phases of the F-actin counterion correlations. The microscopic mechanism involves a
are chosen, a large number of defects will occur for the native coupled mode between a counterion CDW and the polyelectro-
13兾6 symmetry, in which sd-1s of adjacent filaments are in very lyte twist. (The counterion CDW cannot exist without the
close proximity to one another, which is electrostatically and polyelectrolyte twist and vice versa.) Because most polyelectro-
sterically unfavorable. Given that F-actin is a helix, defects of this lytes are more flexible than F-actin, variations of these coupled
kind are unavoidable, unless the helix was completely unwound. cooperative distortions that facilitate counterion ordering and
However, the spatial frequency of these defects is dramatically like-charge attraction are likely to occur in a wide range of
reduced if the F-actin filament is overtwisted to a 36兾17 sym- polyelectrolyte systems. This molecular mechanism is analogous
metry. The 36兾17 symmetry provides long regions with evenly to the formation of polarons in ionic solids, in which an electron
spaced negative-charge patches with a period equal to twice drags a ‘‘cloud’’ of longitudinal optic phonons and acquires a
the monomer spacing on a single filament, d ⫽ 57.4 Å, which is large effective mass (22, 23). In the present polyelectrolyte
quite close to 59.8 Å, our measured period for the 1D CDW of system, fluctuating counterions drag along soft helical distor-
multivalent ions. In fact, this observed organization of multiva- tions of the polyelectrolyte, and consequently freeze into a static
lent ions into a classical CDW may be related to the occurrence correlation, thus providing a transition between the extreme
of these defects. viewpoints of dynamic and static counterion correlations. The
Does optimizing the electrostatic interactions in a single effect is analogous to superconductivity, in which a lattice
interstitial channel necessarily do so for all of the interstitial distortion mediates an attraction between two like-charged
channels? The answer to this question leads to one other particles; in the polyelectrolyte system, charged particles mediate
important advantage in changing the actin twist state from 13兾6 an attraction between two like-charged distorted lattices. In a

APPLIED PHYSICAL
to 36兾17. There are 36 monomers in the crystallographic repeat broader context, similar collective compromises between F-actin

SCIENCES
of twisted F-actin in the 36兾17 configuration. Thirty-six is twist and crosslinking of F-actin by actin binding proteins may
divisible by 6, whereas 13, the number of monomers in the native influence the hierarchy of existing interactions (24) and have
crystallographic repeat of F-actin, is not. In the limit of 6-fold important thermodynamic consequences for cytoskeletal regu-
coordination of 36兾17 F-actin rods into a columnar arrangement, lation. Another general issue is the role of ion multivalence in
it can be shown that all of the interstitial channels can be like-charge attraction, because different polyelectrolytes require
optimized simultaneously, because all channels within a periodic ions of different valences to condense. An experimentally mo-
unit cell can be optimized (only one of the two possible channel tivated criterion for the degree of ion multivalence required for
types are shown in Fig. 4), whereas the same is not true for 13兾6 generating attractions has been recently proposed (25), and a
F-actin. systematic study of collective counterion dynamics for different
The torsional modulus of F-actin has been measured, and the ion valences is under way.
energetic cost of twisting the filament by 3.8° is relatively small
(Etwist ⫽ 1⁄2␬ ⌬␪2d⫺1 ⫽ 1.6 kT, ␬ ⬇ 8 ⫻ 10⫺26 Nm2) (19). By using We thank K. Schweizer and K. Schulten for helpful discussions, and J.H.
for technical assistance. This work is based on work supported by
the measured value for the torsional modulus and the equipar- National Science Foundation Grant NSF-DMR-0071761, the Beckman
tition theorem, the rms twist of F-actin at room temperature just Young Investigator Program, Grant 00G0 from the Cystic Fibrosis
from thermal fluctuations is ⬇3°. Foundation, and the U.S. Department of Energy, Division of Materials
The possibility of coupling between the CDWs in neighboring Sciences under Award DEF G02-91ER45439, through the Frederick
tunnels deserves comment. There is not sufficient dynamic range Seitz Materials Research Laboratory at the University of Illinois.

1. Kirkwood, J. G. & Shumaker, J. B. (1952) Proc. Natl. Acad. Sci. USA 38, 15. Al-Khayat, H. A., Yagi, N. & Squire, J. M. (1995) J. Mol. Biol. 252, 611–632.
863–871. 16. Tibbits, T. T. & Caspar, D. L. D. (1993) Acta Crystallogr. A 49, 532–545.
2. Oosawa, F. (1971) Polyelectrolytes (Dekker, New York). 17. Egelman, E. H., Francis, N. & DeRosier, D. J. (1982) Nature 298, 131–135.
3. Ha, B. Y. & Liu, A. J. (1997) Phys. Rev. Lett. 78, 1289–1292. 18. Aebi, U., Millonig, R., Salvo, H. & Engel, A. (1986) Ann. N.Y. Acad. Sci. 483,
4. Podgornik, R. & Parsegian, V. A. (1998) Phys. Rev. Lett. 80, 1560–1563. 100–119.
5. Lyubartsev, A. P., Tang, J. X., Janmey, P. A. & Nordenskiold, L. (1998) Phys. 19. Tsuda, Y., Yasutake, H., Ishijima, A. & Yanagida, T. (1996) Proc. Natl. Acad.
Rev. Lett. 81, 5465–5468. Sci. USA 93, 12937–12942.
6. Stevens, M. J. (1999) Phys. Rev. Lett. 82, 101–104. 20. Chaikin, P. M. & Lubensky, T. C. (1995) Principles of Condensed Matter Physics
7. Rouzina, I. & Bloomfield, V. A. (1996) J. Phys. Chem. 100, 9977–9989.
(Cambridge Univ. Press, New York).
8. Shklovskii, B. I. (1999) Phys. Rev. Lett. 82, 3268–3271.
21. Wong, G. C. L., de Jeu, W. H., Shao, H., Liang, K. S. & Zentel, R. (1997) Nature
9. Kornyshev, A. A. & Leiken, S. (1999) Phys. Rev. Lett. 82, 4138–4141.
389, 576–579.
10. Gronbech-Jensen, N., Mashl, R. J., Bruinsma, R. F. & Gelbart, W. M. (1997)
Phys. Rev. Lett. 78, 2477–2480. 22. Kittel, C. (1966) Introduction to Solid State Physics (Wiley, New York).
11. Manning, G. S. (1978) Q. Rev. Biophys. 2, 179–246. 23. Feynman, R. P. (1972) Statistical Mechanics: A Set of Lectures (Benjamin,
12. Lodish, H., Berk, A., Zipursky, L. S., Matsudaira, P., Baltimore, D., Darnell, Reading, MA).
J. & Zipursky, L. (2000) Molecular Cell Biology (Freeman, New York), 4th ed. 24. McGough, A., Pope, B., Chiu, W. & Weeds, A. (1997) J. Cell. Biol. 138,
13. Sheterline, P., Clayton, J. & Sparrow, J. C. (1998) Actin (Oxford Univ. Press, 771–781.
New York). 25. Butler, J., Angelini, T. E., Tang, J. X. & Wong, G. C. L. (2003) Phys. Rev. Lett.,
14. Holmes, K. C., Popp, D., Gebhard, W. & Kabsch, W. (1990) Nature 347, 44–49. in press.

Angelini et al. PNAS 兩 July 22, 2003 兩 vol. 100 兩 no. 15 兩 8637