Cell, Vol. 98, 537±546, August 20, 1999, Copyright 1999 by Cell Press
Structure of the ?-Actinin Rod: Molecular Basis
for Cross-Linking of Actin Filaments
?-Actinin is composed of an amino-terminal actin-
binding regionconsisting oftwocalponinhomology(CH)
domains, a central rod containing four spectrin-like re-
peats, and a calmodulin-like domain at the carboxy ter-
minus (Castresana and Saraste, 1995; Trave et al., 1995;
Davison and Critchley, 1988). It is a dimer composed of
lel mannerto forma rod-shaped molecule with an actin-
binding region at either end (Figure 1C). This arrange-
ment allows ?-actinin to cross-link actin filaments into
tight bundles. Other proteins of the family (spectrin and
dystropin) are composed of the same building blocks
but differ in the number of repeats that separate the
actin-binding regions in the oligomeric structures. The
length of the spacer determines the final structure of
the higher-order cytoskeletal assembly in the cell.
The repeats are characteristic of the entire family and
contain from 100 to 120 residues. They were identified
as homologous repeats in the sequence of ?-spectrin
(Speicher and Marchesi, 1984) and were later discov-
ered in ?-actinin and dystrophin (Davison and Critchley,
1988). The repeats are independent folding units, as
shown by structural studies on a single repeat (Pascual
et al., 1997b) and by unfolding studies using atomic
force microscopy onspectrinand ?-actininrepeats (Rief
et al., 1999).
The crystal structure of the 14th repeat from the Dro-
sophila melanogaster ?-spectrin (Yan et al., 1993) and
solution structure of the 16th repeat from the chicken
brain ?-spectrin (Pascual et al., 1997b) show that the
repeat is an antiparallel triple-helical structure. The heli-
ces within the repeats show the heptad sequence pat-
tern, which is commonly found in extended ?-helical
structures (McLachlan and Stewart, 1975). When the
positions are canonically labeled from a to g, the resi-
dues a and d are generally hydrophobic and located on
the inward-facing surface of the helix. These residues
form the core of the domain and stabilize its fold by
hydrophobic interactions. They are well conserved in
the sequence alignments of homologous repeats in
spectrins and ?-actinins (Pascual et al., 1997a). Yan et
al. (1993) have proposed a model in which successive
repeats are connected by a long continuous ? helix. The
structure of two consecutive repeats from the chicken
?-spectrin confirms the presence of a long helical con-
nection between the repeats that may form a structural
basis for the flexibility of the spectrin molecule (Grum
et al., 1999 [this issue of Cell]).
Dimerization of ?-actinin and spectrin is thought to
be largely due to the contacts between the subunits
that are mediated by the repeats (Imamura et al., 1988;
Kahana and Gratzer, 1991; Speicher et al., 1992; Viel
and Branton, 1994). In spectrin, a nucleation site for the
formation of the ?/? heterodimer has been found near
one end of the molecule, and dimerization is proposed
to proceed ina zipper-likefashion(Speicheretal., 1992).
The arrangement of the spectrin-like repeats in the
?-actinin homodimer (hereafter referred to as R1 to R4)
is often depicted in an aligned assembly (Figure 1CI).
However, a staggered model has been proposed based
Kristina Djinovic Â-Carugo, Paul Young,
Mathias Gautel, and Matti Saraste*
Structural Biology Programme
European Molecular Biology Laboratory
We have determined the crystal structure of the two
central repeats in the ?-actinin rod at 2.5 AÊresolution.
The repeats are connected by a helical linker and form
a symmetric, antiparallel dimer in which the repeats
are aligned ratherthanstaggered.Using this structure,
which reveals the structural principle that governs the
architecture of ?-actinin, we have devised a plausible
model of the entire ?-actinin rod. The electrostatic
properties explain how the two ?-actinin subunits as-
semble in an antiparallel fashion, placing the actin-
binding sites at both ends of the rod. This molecular
architecture results in a protein that is able to form
cross-links between actin filaments.
?-Actinin is a ubiquitously expressed protein that is re-
garded as theancestralmoleculewithinafamily ofactin-
binding proteins that includes spectrin, dystrophin, and
utrophin (Blanchard et al., 1989; Pascual et al., 1997a).
Muscle and nonmuscle isoforms of ?-actinin have been
characterized (Blanchard et al., 1989). In general, the
nonmuscle isoforms bind to F-actin in a calcium-sensi-
tive manner, whereas actin binding of the muscle iso-
forms is not controlled by calcium. In skeletal and car-
diac muscle, ?-actinin is found in the Z disk, where
it cross-links antiparallel actin filaments from adjacent
sarcomeres (Figure 1A). Interactions between ?-actinin
and titin may play a role in controlling Z disk assembly
(Ohtsuka et al., 1997; Young et al., 1998). In nonmuscle
cells, ?-actinin is involved in the organization of the
structures such as zonula adherens and tight junctions.
In cultured fibroblasts, it is localized at focal contacts
or along stress fibers (see Blanchard et al., 1989 and
references therein). In these structures, ?-actinin not
only cross-links actin filaments but also interacts with
numerous cytoskeletal and membrane-associated pro-
teins. Itmay link the actincytoskeletonto the membrane
either directly, via interactions with transmembrane re-
ceptors such as integrins, ICAMs, L-selectin, and the
NMDA receptor (Otey et al., 1990; Carpen et al., 1992;
Pavalko et al., 1995; Heiska et al., 1996; Wyszynski et
al., 1997), or indirectly, via cytoskeletal proteins such
as vinculin (Wachsstock et al., 1987; Figure 1B).
*To whom correspondence should be addressed (e-mail: saraste@
on projection images from two-dimensional crystals of
?-actinin. In this model, either R1 or R4 does not pair
with any repeat of the opposing monomer (Figures 1CII
and 1CIII; Taylorand Taylor, 1993). Studies on the dimer-
izationofexpressed fragments containing eitherthreeor
four repeats strongly support the aligned arrangement
(Flood et al., 1995, 1997), whereas chemical cross-link-
ing has not distinguished between the aligned and stag-
gered models (Imamura et al., 1988).
Here we present the structure of the two central re-
peats (R2R3) of the muscle ?-actinin, as determined by
X-ray crystallography at 2.5 AÊresolution. The crystal
structure shows that the two-repeat segments form a
symmetric, antiparallel dimer and reveals the structural
principle that governs the architecture of the entire
?-actinin rod and gives the protein the ability to form
cross-links between actin filaments.
Results and Discussion
The central repeats R2R3, corresponding to residues
371±637 of the human skeletal muscle ?-actinin 2, crys-
tallized in space group P6522 with one molecule per
asymmetric unit. Experimental phases were determined
using multiwavelength anomalous dispersion (MAD) of
the mercury derivative (Table 1). The solvent-flattened
electron density map was readily interpretable and al-
lowed positioning of most of the backbone of R2R3 and
about65% ofthe side chains. This modelwas submitted
to several cycles of refinement and model building. The
finalrefinedmodelhas afreeR valueof31.0%,aconven-
tional crystallographic R value of 22.8% (using all data
to 2.5 AÊ), and consists of 248 amino acid residues and
123 water molecules. The number of water molecules
is in agreement with the statistical analysis on experi-
mentally located solvent molecules in protein crystal
structures (Carugo and Bordo, 1999).
Figure 1. Domain Structure of ?-Actinin and Its Function in the Sar-
comeric Z Disk and in Focal Contacts
(A) In the muscle Z disk, ?-actinin (?A) cross-links antiparallel actin
filaments from adjacent sarcomeres. Titin acts as a molecular ruler
forthe sarcomere and interacts with two different parts of ?-actinin,
atthecenterofthe?-actininrod andwiththecalmodulin-like domain
(Gregorio et al., 1999). These interactions may play a role in control-
ling the thickness of the Z disk (Young et al., 1998).
(B) A simplified representation of a focal contact showing ?-actinin
linking the actin cytoskeleton to membrane-associated structures.
Focal contacts are points where cultured cells are attached tightly
to the extracellular matrix via transmembrane receptors such as
integrins (? and ? in figure) (Burridge and Chrzanowska-Wodnicka,
1996). ?-Actinin has been shown to interact with ? integrins (Otey
et al., 1990) as well as with the focal contact components vinculin
(Wachsstock et al., 1987) and zyxin (Crawford et al., 1992). Thus, it
may connect integrins to actin filaments either directly or indirectly,
involving proteins such as talin(Horwitz et al., 1986) ortensin (Beck-
(C) Domain structure of ?-actinin showing aligned and staggered
models for the dimer. ABD, actin-binding domain; R1, R2, R3, and
R4, repeats; C, C-terminal calmodulin-like domain. The color
scheme for the repeats is maintained throughout the paper. (I), an
aligned arrangement where R1 and R2 are paired with R4 and R3,
respectively, ofthe opposing monomer; (II)and (III), alternative stag-
gered arrangements where either R1 or R4 is not paired with any
repeat of the opposing monomer.
Structure of the Double Repeat
The three-dimensionalstructure consists oftwo repeats
connected by a helical linker (Figure 2A). The molecule
has an elongated shape with a length of 130 AÊand a
diameter of about 20 AÊ. The fold of each repeating unit
is determined by three ? helices in a coiled-coil assem-
bly. In the following, we shall refer to the helices of the
first repeat as 1, 2, and 3 and the helices of the second
repeat as 1?, 2?, and 3?.
The two repeats are structurally very similar. The su-
perpositionofR2 and R3 results ina 1.36 AÊrms distance
of 84 equivalent C? atoms (Figure 2B), and the rms
distances between the C? atoms of repeat R16 of
?-spectrin (Pascual et al., 1997b) and R2 and R3 are
1.37 AÊand 1.40 AÊ(on 64 C? atoms), respectively. The
major structural difference between R2 and R3 resides
in the loop connecting the second and the third helix
(loop 2-3/2?-3?). In R2, helices 2 and 3 are connected
by a ? turn, whereas the corresponding region in R3 is
much longer, with residues 209±215 protruding toward
the helix 1? (for residue numbering, see Figure 6A). The
loop conformation is locked in this position by several
interactions with helix 1. The second prominent differ-
ence is found in the C-terminal part of the last helix. In
R2, helix 3, which connects to R3 through the helical
linker, has a linear conformation stabilized by interac-
tions with R3. The end of helix 3? in R3 is bent toward
the core of the protein, but its conformation may be
more linear when the R4 repeat is present (see below).
Apart from the hydrophobic packing in the core, elec-
trostatic interactions also contribute to the stabilization
Structure of ?-Actinin Rod
Table 1. Data Collection and Refinement Statistics
a ? b ? 60.59
c ? 390.47
a ? b ? 60.05
c ? 390.81
Figure of merit (55.0±2.9 AÊ)c
No. of reflections
Bond lengths (AÊ)
Bond angles (?)
Overall B value (AÊ2)
aNumbers in parentheses refer to the last resolution shell.
bRmerge? ?|Ii? ?I?|/? Ii, where Iiis the intensity of an individual reflection and ?I? is the mean intensity of that reflection.
cFigure of merit ? ?|?P(?)ei?/?P(?)|?, where ? is the phase and P(?) is the phase probability distribution.
dR factor ? ? |Fo? Fc|/?Fo.
eRfreeis the cross-validation R factor computed for the test set of reflections (8% of the total) that are omitted in the refinement process.
of the repeat fold. Salt bridges and hydrogen bonds
between oppositely charged residues can be grouped
into two classes: intrahelical interactions between resi-
dues spaced at i, i ? 3, ori ? 4 intervals, and interhelical
contacts that contribute to the tertiary structure of each
repeat. Both types of interactions can be observed in
?-helical structures (Lyu et al., 1992; Kohn et al., 1998)
and are known to stabilize ? helices and coiled-coil
assemblies. The residues involved in these interactions
are notstrictly conserved eitherinR2orR3 norgenerally
in the ?-actinin repeats. However, functionally analo-
gous contacts are observed when R2 and R3 structures
Figure 2. Overall Structure of R2R3 and Structural Comparison of the Repeats
(A) R2 is colored blue, R3 is green, and the linker segment is red. All ribbon diagrams were generated using the programs MOLSCRIPT (Kraulis,
1991) and Raster3D (Merritt and Murphy, 1994).
(B) Stereoview of the superposition of the helices in R2 and R3. Coloring is as depicted above.
Figure 3. The Connecting Linker
Close-up of interactions between R2, R3, and the linker. The protein
backbone is shown as a ribbon, and amino acid residues are drawn
in a ball-and-stick representation. R2 is colored blue, R3 is green,
and the linker is red.
Figure 4. Sedimentation Equilibrium
Analytical ultracentrifugation was carried out as described in the
Experimental Procedures. The figure shows absorbance at 293 nm
versus ?r2, where r is the radial position. The experimental data is
shown as open circles and the fitted data as a continuous line.
The bottom shows the residuals (open circles) expressed as the
difference between the experimental and the fitted data.
are compared. Interestingly, only one such ion pair in-
volves residues coming from the two consecutive re-
peats (Arg51±Glu125)and joins helix 2ofR2to the linker,
connecting the two repeats (Figure 3).
The boundaries of repeats in ?-actinin have been exten-
sively studied by limited proteolysis. According to our
structure, the N-terminal boundaries of proteolytically
stable constructs, as defined by Gilmore et al. (1994)
for R3 and R4, are located within the third helix of the
preceding repeat. This may indicate that the C-terminal
helix of the repeat, the N-terminal helix of the following
repeat, and the helical linkerforma stable, folded struc-
The linker region between the two repeats can be
determined by an analysis of the gradient of atomic
displacement parameters. A sharp and suddenincrease
of the gradient of atomic displacement parameters is
observed between residues Met-120 and Glu-125 at the
three-dimensionalinterface ofthe two repeats thatmeet
along the connecting helix at residue 122. The segment
120±126 is flanked by the structural elements of the two
consecutive repeats, in particular by loop 1-2 and helix
2 and loop 2?-3? and helix 2?. One side of the linkerhelix
is exposed to the solvent (Figure 3).
We define the ?-helical section, which is colored red
in Figures 2A and 3, as the linker between the two con-
secutive repeats. The interactions within the linker re-
gion are centered at the hydrophobic residues Leu-123
and Leu-124, which make van der Waals contacts both
to the residues of R2 (Leu-47) and R3 (Ile-206 and Tyr-
202; the latter also contacts the side chain of Met-120).
Apart from this hydrophobic cluster, additional polar
interactions between the residue pairs Glu43±Asn203 (a
side chain±to±side chain hydrogen bond) and Arg51±
Glu125 (a salt bridge) stabilize the repeat interface.
These residues are highly conserved in ?-actinins
(Blanchard et al., 1989), suggesting that this interaction
geometry between repeats R2 and R3 is a structural
feature of the family. An exception is ?-actinin from
Dictyostelium discoideum, in which the 2?-3? loop is
seven residues shorter and must adopt a different local
The relativeorientationofthe tworepeats is described
by a translation ofabout 55 AÊalong the connecting helix
and a rotationby approximately 100 degrees around the
same axis. A proline residue at position 140 induces a
bend of 21 degrees in the connecting helix. The pres-
ence of a proline at this position is a conserved feature
within the ?-actinin family, but proline is not preserved
in all repeats within the ?-actinin molecule. Thus, the
connecting helices between other repeats could adopt
variable conformations in this region and have different
degrees of bending.
The Antiparallel Dimer
We have analyzed the oligomeric state of the R2R3 con-
structwithseveraltechniques and at various concentra-
tions. Yeast two-hybrid analysis, size-exclusion chro-
matography, and analytical ultracentrifugation clearly
indicate that the R2R3 construct forms a stable dimer
in solution at concentrations between 1±50 mg/ml. The
analytical centrifugation data (Figure 4) show that the
sampleis a homogeneous dimerwitha molecularweight
of60.8kDa, ingood agreementwiththetheoreticalvalue
of 58.5 kDa. Kdfor the monomer/dimer dissociation can
be estimated to be 2 ?M. The crystal structure also
shows a dimer formed by a crystallographic 2-fold axis,
which is perpendicular to the long molecular axis and
centered at the middle ofthe dimer. The dimeris assem-
bled in an aligned, antiparallel manner (Figure 5A). The
two monomers are in contact along the whole length of
Structure of ?-Actinin Rod
Figure 5. Structure of the R2R3 Dimer and the Interface
(A) Ribbon diagram of the antiparallel R2R3 dimer. R2 is blue, R3 is green, the linker is red, and the structural elements involved in the dimer
interface are yellow.
(B) Dimer interface. The protein backbone is shown as a ribbon, and the residues involved in the dimer interface are depicted as red balls
for R2 and yellow balls for the opposing R3, centered on the C? atom of the residues.
(C) Electrostatic surface potential of the R2R3 dimer interface. Surfaces are colored by electrostatic potential. Positive regions are depicted
in blue and negative in red. The second subunit (right) was rotated by 180 degrees to show charge complementarity in the dimer interface.
Figure was generated with GRASP (Nicholls et al., 1991).
the molecule, and 9.2% (1460 AÊ2) of the monomer sur-
face is buried upon dimer formation. This value is in
good agreement with corresponding values for stable
homodimers (J ones and Thornton, 1996).
In the dimer interface, helix 1? and loop 1?-2? of R3
face the groove formed by helices 1 and 2 of R2 in the
opposing monomer (Figure 5B). The C-terminal region
of helix 1? and the subsequent loop 1?-2? of R3 interact
withthe N-terminalpart ofhelix 1 in R2 (top). The central
part of the helix 1? interacts with both helices 1 and 2
(center), whereas the N-terminalpartofhelix 1?, together
with the protruding loop 2?-3?, contacts only helix 2 of
R2 (bottom). These interactions, which are lined along
the longitudinal direction of the R2R3 structure, are re-
peated in the second half of the dimer due to the 2-fold
symmetry axis perpendicularto the long molecularaxis.
In total, 38 residues per monomer can be considered
to be at the dimer interface, as judged by the decrease
in solvent-accessible area upon dimer formation.
The contacts at the top of the dimer consist of a
hydrophobic cluster and specific polar interactions.
Loop 1?-2? (R3) is lined above helix 1 (R2) and stabilized
inthis positionby hydrophobic interactions totheunder-
lying N-terminalpartofhelix 1. Additionally, a saltbridge
and a main chain±to±main chain hydrogen bond are
involved in contacts between residues from loop 1?-2?
and helix 1 (Figure 5B). In the center, the interface is
formed by two solvent molecules that establisha hydro-
gen bonding network connecting Glu-150 (helix 1?) to
His-26 (helix 1) and His-67 (helix 2) and to the carbonyl
oxygen of Lys-22 (helix 1). These interactions join two
? helices in R2 and one helix in R3 of the other subunit.
Additionally, a salt bridge between Arg-70 (helix 2) and
Glu-150 (helix 1?) contributes to the stability of the
?-actinin dimer in this region.
At the bottom end of the R2±R3 interface, helices 2
and 1? interact through a hydrogen bonding network
bridged by a solvent molecule and through van der
Waals interactions betweentwo smallhydrophobic resi-
dues. The interactions between the two monomers con-
tinue throughcontacts betweenthe helix 2 residues and
the protruding part of the loop 2?-3?, involving residues
211±214. Several direct polar interactions, as well as
solvent-mediated hydrogen bonds involving side chain
and main chain atoms, are found in this region. More-
over, a small hydrophobic cluster centered around Tyr-
212 (loop 2?-3?) is formed upon dimerization. Finally, a
hydrogen bond between the two symmetry-related Thr-
48 residues in the opposing subunits terminates the
dimer interface at the proximity of the 2-fold axis.
The majority of interactions in the dimer interface are
polar, and some ofthemare modulated by solventmole-
cules that assist the continuation of the hydrogen bond-
ing network connecting thesubunits.Thepredominantly
polar character of the contacts between the two sub-
units is likely to be the basis of salt-induced conforma-
tionalchanges of?-actininobserved by electronmicros-
copy and viscosity measurements (Kuroda et al., 1994;
Winkler et al., 1997).
Analysis of the electrostatic surface potential of the
dimer interface (Figure 5C) shows complementarity be-
tween the interacting surfaces in terms of productive
electrostatic interactions between the two monomers.
Ina staggered model(Figures 1CIIand 1CIII), the juxtapo-
sition of R2 or R3 would result in a number of repulsive
interactions between equally charged residues and pre-
vent the dimerization.
The structural elements involved in the interface lead
dimer (Figure 1CI). Most of the residues in the interface
Figure 6. Model of the ?-Actinin Rod
(A) Sequence alignment of the ?-actinin repeats used in modeling of repeats R1 and R4. Residue numbers for the full-length molecule and
those of the construct used for the crystal structure are indicated at the edges and above the alignment, respectively. The ? helices seen in
the crystal structure are depicted as bars (blue for R2 and green for R3). The C termini of R1 and R3 and the N termini of R2 and R4,
respectively, overlap due to the modeling procedure of R1 and R4. The overlapping residues shown in italic were used to assemble the model
of the rod. The figure was generated with ALSCRIPT (Barton, 1993).
(B) Ribbon diagram of ?-actinin rod viewed in two orientations related by a 65 degree rotation around the long molecular axis through the
central 2-fold axis. R1 is colored violet, R2 is blue, R3 is green, and R4 is yellow.
(C) Electrostatic surface potential of one R1±R4 subunit generated with GRASP (Nicholls et al., 1991). Positively charged surface is colored
blue and negatively charged red. The second subunit is white in a wormlike representation. Surfaces corresponding to R1 and R4 are marked.
are well conserved within the ?-actinin sequences
(Blanchard et al., 1989). This is expected, as disruptive
mutations affecting residues involved in the dimerinter-
face would have a double effect on its stability due to
the 2-fold symmetry.
rotation of about 120 degrees around the connecting
helix, together with a translation of approximately 55 AÊ
along the same axis. Similar rotations and translations
(125 degrees and 60 AÊ) characterize spatial relations
between repeats R3 and R4. The model of the mono-
meric rod deviates by 20 degrees from linearity.
The dimer is around 240 AÊlong and 40 AÊwide and
displays a twist of 73 degrees. These dimensions are in
a good agreement with electron microscopic data on
the intact ?-actinin (22.5 ? 2.1 nm; Flood et al., 1995)
and on the polypeptide released from ?-actinin by ther-
molytic digestion(22.5nm?5.9nm;Winkleretal., 1997).
3380 AÊ2(12%) of the solvent-accessible area in the
monomer is involved in the dimer interface, which con-
tinues throughout the long molecular axis and involves
all four repeats. The organization of the ?-actinin rod
brings helix 3 of R4 and helix 3 of the opposing R1 into
proximity. The interactions between these two moieties
are based on polar contacts. Additional polar interac-
tions between repeats R1 and R4 involve the two pairs
In order to further elucidate the molecular basis for
cross-linking of actin filaments by ?-actinin, we have
constructed a model of the ?-actinin rod, comprising
repeats R1±R4. A homology model of the monomeric
rod was built using the three-dimensional structures of
domains R2 and R3 as templates for modeling repeats
R4 and R1, respectively, as described underExperimen-
tal Procedures, using the sequence alignment shown in
Figure 6A. The antiparallel dimer (Figure 6B) was con-
structed by application of the crystallographic 2-fold
axis, which relates the two monomers in the crystal
structure of R2R3. The relative orientation of repeats R1
and R2 in the monomeric rod can be described by a
Structure of ?-Actinin Rod
of structural elements: helix 3 of R4 and helix 2 of R1,
and loop 2-3 of R4 and helix 3 of R1.
A number of experiments employing different meth-
ods have been used to address the question of the
intersubunit interactions within the ?-actinin homodi-
mer.EM studies onchickengizzard ?-actinininnegative
stain (Taylor and Taylor, 1993) and under different salt
conditions (Winkleretal., 1997)supportstaggered mod-
els (Figures 1CIIand 1CIII) and suggest conformational
changes induced by a variation of ionic strength. Also,
EM experiments on fragments containing repeats 1±4,
1±3, and 2±4 canbe interpreted as being consistent with
a staggered model (Flood et al., 1995). However, none
of these EM studies are at sufficiently high resolution
to allow unambiguous identification of the individual re-
peats, and as pointed out by Flood et al. (1995), it is
not clear whether the ?-actinin dimer retains its native
structure under the harsh conditions used for sample
preparation in some EM experiments.
Biochemical data based on cross-linking and analy-
sis ofassociationby sedimentationequilibriumofdiffer-
ent constructs containing repeats R1±R4, R2±R4, and
R1±R3 have shownthat inorderto generate a maximally
stable antiparallel dimer, all four repeats are required.
This suggests that they all contribute to the dimer inter-
face and indicates their aligned pairing. In addition, our
yeast two-hybrid experiments on double repeat con-
structs indicate that R1R2 interacts with R3R4 and that
R2R3 interacts with itself, consistent with an aligned
model. No interactions are observed between R2R3 and
either R1R2 or R3R4, arguing against a stagger within
the rod in either direction (Figure 1C; P. Y., unpublished
In our model, the repeats of opposing subunits in the
dimer (Figure 6B) interact in a pairwise manner corre-
sponding to the aligned model. The model is in good
agreement with the biochemical studies mentioned
above. The isolated rod dimer is extremely stable (Ima-
mura etal., 1988; Kahana and Gratzer, 1991; Flood etal.,
1995), suggesting that the N- and C-terminal domains
of ?-actinin do not make a dominant contribution to
dimerization. Thus, we expect that the arrangements of
the repeats in our model closely resemble that of the
full-lengthmolecule.Giventhe highsequence conserva-
tion in the ?-actinin rod between the muscle and non-
muscle isoforms (Beggs et al., 1992), it is almost certain
that the aligned model holds true for both.
The electrostatic potential of the monomeric rod (Fig-
ure 6C) shows a pronounced polarity of the molecule,
with a positive charge at the N terminus and a negative
charge at the C terminus. In the aligned, antiparallel
assembly of subunits, oppositely charged repeats are
brought in optimal juxtaposition, stabilizing the dimer
interface. Inspection of the level of amino acid identity
between the ?-actinin repeats (Figure 6A) shows that
R1 and R4 have diverged further during their evolution
with respect to R2 and R3. This is expected, since their
prominent charge complementarity is required for the
optimal fit between subunits. Taken together, the analy-
sis of the dimer interface along with biochemical data
strongly supports the aligned, antiparallel model for the
?-Actinin links diverse cellular structures to the actin
cytoskeleton in many different cell types. The binding
sites ofmany ligands have beenmapped to the ?-actinin
rod. In the Z disk of muscle, these partner proteins in-
clude titinand a LIM protein, whichboth appearto inter-
act with the ?-actinin rod (Xia et al., 1997; Young et al.,
1998). The rod region also binds to the cytoplasmic tail
of the NMDA receptor (Wyszynski et al., 1997). R3 of
bothskeletal and nonmuscle ?-actininmay interact with
the Rho-activated protein kinase PKN (Mukai et al.,
1997). The interactions between ?-actinin and the cyto-
plasmic domains of the transmembrane receptors ?1
integrin and L-selectin have been mapped to the rod
(Otey et al., 1990; Pavalko et al., 1995). Our structure
may help to study these interactions in more detail.
Molecular architecture of the ?-actinin dimer forms
the basis of its primary function, cross-linking of actin
filaments. The repeats of the rod define the elongated
shape ofthe moleculeas wellas theantiparallelassocia-
tion of the subunits that places the functional domains
(calmodulin-like domainand actin-binding region) at the
ends of the molecule. In skeletal and cardiac muscle,
?-actinin cross-links antiparallel actin filaments coming
fromadjacent sarcomeres (Figure 1). The symmetry axis
at the centerof the rod can easily lead to the antiparallel
arrangement of actin-binding sites in the dimer, which
is complementary to the antiparallel orientation of actin
filaments inthe Z disk.Innonmuscle and smoothmuscle
cells, actin filaments are not part of an ordered lattice
peculiar to the Z disk and can assume variable orienta-
tions. In orderto accomplish actin bundling underthese
conditions, the actin-binding domain must be able to
change its orientation via a built-in flexibility in the neck
connecting it to the rod. The fact that the actin-binding
fragment canbe isolated fromthe rod by limited proteo-
lytic digestion (Imamura et al., 1988) indicates such an
inherent flexibility of this neck. A principal feature that
distinguishes the muscle and nonmuscle ?-actinins is
the ability of the latter to bind actin filaments in a cal-
cium-sensitive manner (Landon et al., 1985). Calcium
concentrations higher than 10?7M abolish binding. The
antiparallel orientation of the subunits makes it possible
of one subunit sterically prevents actin binding of the
neighboring subunit via a conformational change. Eluci-
dationofthis mechanismwillrequire thestructure deter-
mination of the entire ?-actinin molecule.
Preparation and Crystallization of the R2R3 Construct
A DNA fragment corresponding to amino acids 371 to 637 of human
skeletal muscle ?-actinin 2 (ACTN2; EMBL database M86406) was
amplified froma humancardiac cDNA library (Clontech)by polymer-
ase chain reaction. This region contains the spectrin-like repeats 2
and 3 (R2R3) as determined by Gilmore et al. (1994) for chicken
cytoskeletal?-actinin.This fragmentwas clonedintoapET 8c vector
(Novagen) modified such that the R2R3 polypeptide contains the
additional N-terminal amino acid sequence MHHHHHHSTEN
LYFQGSS when expressed in Escherichia coli BL21[DE3]. Induction
of expression was carried out at 37?C using 0.2 mM IPTG. Absence
of mutations was verified by DNA sequencing. Initial purification
was carried out on an Ni-NTA agarose column (Qiagen) followed by
removal of the N-terminal tag using TEV protease (GIBCO-BRL)
(Parks et al., 1994). This results in an ?-actinin R2R3 polypeptide
containing three additionalN-terminalresidues, GSS. This untagged
polypeptide was collected in the flowthrough fraction of a second
Ni-NTA column and further purified by ion exchange on a MonoQ
column (Pharmacia). The protein was then dialyzed to 20 mM Tris-
HCl (pH 7.4), 1 mM DTT and concentrated to 150 mg/ml forcrystalli-
zation. Crystals were grown using vapor diffusion at 4?C or 17?C by
mixing the protein in a 1:1 volume ratio with a solution containing
26% PEG 400 (Fluka), 100 mM MgSO4, and 100 mM HEPES or
Tris-HCl (pH 7.5±8.0). Crystals were hexagonal bipyrimids and had
dimensions ofupto0.5mm.Thehighproteinconcentrationis neces-
sary to obtain crystals under these conditions.
with the 48.6 AÊ2overall thermal factor obtained from Wilson scaling
of the diffraction data (CCP4).
Sedimentation equilibrium studies of R2R3 were carried out on a
Beckman Optima XL-A analytical ultracentrifuge using an An50Ti
rotor at a protein concentration of 1 mg/ml in a buffer containing
20 mM Tris-Cl (pH 7.6) and 1 mM DTT. Centrifugation was carried
out at 4?C and 11,000 rpm for 24 hr. Absorbance was monitored at
293 nm. The data were analyzed using the Ultrascan Version 2.98
package (B. Demeler, The University ofTexas HealthScience Center
at San Antonio).
Construction of Homology Models of R1 and R4
and ?-Actinin Rod
Sequence alignments were produced with CLUSTAL W-1.6 (Higgins
et al., 1991) and edited with the SEQLAB program of the GCG pack-
age (Wisconsin Package Version 9.1, Genetics Computer Group
[GCG], Madison, WI). The homology modeling of R1 and R4 repeats
was carried out by first obtaining a reliable sequence alignment of
repeats 1±4 of human skeletal muscle ?-actinin 2 (Figure 6A). The
R2 and R3 sequences were aligned based on the superposition of
the C? coordinates using program SUPERIMPOSE (Diederichs,
1995) to provide the best overall structural comparison. After this,
a profile alignment with CLUSTAL W was performed between two
groups of aligned amino acid sequences: (1) R2 and R3, and (2)
multiple sequence alignment of spectrin, utrophin, and dystrophin
repeats (Winder et al., 1995). Alignment 1 contained secondary
structure assignments for gap penalty mask. This step produced
the final profile against which R1 and R4 sequences were aligned
(Figure 6A). As R1 and R3 repeats both have an insertion between
helices 2 and 3, R3 structure was used as the template for R1
modeling, whereas R2 was employed in modeling of repeat R4.
Modeling was performed with the program suite MODELLER (Sali
and Blundell, 1993). The monomeric rod of ?-actininwas assembled
by structuralsuperpositionofoverlapping N-terminaland C-terminal
regions (12 amino acid residues) of the R1 model and R2, and R3
and the R4 model (shown in italics in Figure 6A). This model was
subjected to energy minimization using the X-PLOR program
(Bru Ènger, 1992) in order to release steric strain introduced during
the model-building process. Antiparallel dimer was produced by
application of symmetry operator 1 ? X ? Y, Y, ?Z ? 1/2 on the
monomer and submitted to energy minimization by X-PLOR. The
model of ?-actinin rod has a root-mean-square deviation of 0.012 AÊ
from ideal bond lengths and 2.2? from ideal bond angles.
Data Collection and Structure Determination
Crystals were harvested in a stabilization buffer containing 30%
PEG 400, 100 mM MgSO4, and 100 mM HEPES and frozen in a liquid
nitrogen±cooled stream. All data were integrated and reduced with
the DENZO/SCALEPACK programs (Otwinowski and Minor, 1997),
whereas subsequent manipulations of diffraction data were per-
formed using the CCP4 suite of programs (CCP4, 1994). Crystals
formed in space group P6522, with unit cell dimensions of a ? b ?
60.05 AÊ, c ? 390.81 AÊ, and contain one R2R3 molecule in the asym-
metric unit with a solvent content of approximately 65%. All data
were collected at the European Synchrotron Radiation Facility
(ESRF) in Grenoble, France. A native data set to 2.2 AÊresolution
was collected at ID14-3 beamline on an off-line image plate with
dimensions 400 mm? 800 mm using a wavelengthof 0.947 AÊ(Table
1). Derivatives were identified from diffraction data collected at
beamline BM2 on a CCD detector. A mercury derivative was pre-
pared by soaking the crystals in the stabilization buffer with 1 mM
thiomersal for 12 hr.
MAD data were collected to 2.9 AÊresolutionfrommercury-deriva-
tized crystal at three wavelengths: at the dip of the LIII absorption
edge of mercury, and at two energies above the LI and LIII edges,
respectively,using beamlineBM14andMar345image platedetector
(Table 1). A fluorescence scan on derivatized crystal was taken
to optimize data collection parameters. The scaled and reduced
intensity data were converted to amplitudes using TRUNCATE, and
cross-wavelength scaling was performed using SCALEIT of CCP4.
Two mercury sites were identified in the anomalous and dispersive
Patterson maps, using the program RSPS (CCP4). These sites were
further refined with the program PHASES (Furey and Swaminathan,
1997). Additionally, the native data set was included in the phasing
process. The handedness of the heavy atom sites was determined
by inspecting 3.0 AÊresolution MIRAS maps calculated with the
mercury sites in both possible hands and both enantiomorphic
space groups. Clearelectron density forseveral helices in the R2R3
construct defined the hand of the heavy atom sites unambiguously.
Mercury sites were subsequently refined by SHARP (de la Fortelle
and Bricogne, 1997), resulting in a final overall figure of merit of
0.72 (30.0 to 2.9 AÊ). The phases were further improved with solvent
flipping using the program SOLOMON (CCP4), resulting in an elec-
tron density map to which most of the backbone and about 65%
of the side chains were readily positioned using O (J ones et al.,
1991). Due to anisotropic diffraction along the directionperpendicu-
lar to b*, the data to 2.5 AÊresolution were used during refinement.
Refinement of the model by CNS (Bru Ènger et al., 1998) utilized
the native data set (Table 1) with the maximum likelihood target
function (Pannu and Read, 1996). Corrections for the bulk solvent
were applied during the refinement. Weighting schemes (Read,
1986), as implemented inCNS, were used throughout formap calcu-
lations. Several cycles of model building and refinement were per-
formed to produce the final refined model at 2.5 AÊresolution, with
a crystallographic R value of 22.8% and a free R value of 30.2%.
This model consists of residues 371±635 of the human skeletal
muscle ?-actinin 2 and 123 water molecules. PROCHECK (Laskow-
ski et al., 1993) was used to gauge the stereochemical quality of
the final model. The Ramachandran plot has 89.7% of the residues
lying in the most favored and 9.5% in the allowed regions. The two
residues in unfavorable regions have weak density and are located
at the extended N terminus of the construct (GSS). Refined overall
temperature factors of 57.2 AÊ2(54.9 AÊ2for main chain atoms, 59.2 AÊ2
forside chainatoms, and 58.9 AÊ2forwatermolecules)are consistent
The authors wish to thank W. Burmeister, S. Wakatsuki, M. Roth,
and A. Thompson at the ESRF synchrotron facility in Grenoble for
excellent assistance with diffraction experiments. We are grateful
to M. Way forcriticalreading ofthe manuscript;O.Carugo forhelpful
discussions about structure analysis; H. van der Zandt for help with
discussions onsequence alignments.We are gratefulto R.MacDon-
ald and A. Mondrago Ân for communicating results prior to publica-
tion. This work was supported by the Deutsche Forschungsgemein-
schaft grants Ga405/4-1 and Ga405/3-6.
Received J une 16, 1999; revised J uly 13, 1999.
Barton, G.J . (1993). ALSCRIPT: a tool to format multiple sequence
alignments. Prot. Eng. 6, 37±40.
Beckerle, M.C. (1997). Zyxin: zinc fingers at sites of cell adhesion.
Bioessays 19, 949±957.
Beggs, A.H., Byers, T.J ., Knoll, J .H., Boyce, F.M., Bruns, G.A., and
Kunkel, L.M. (1992). Cloning and characterization of two human
skeletal muscle ?-actinin genes located on chromosomes 1 and 11.
J . Biol. Chem. 267, 9281±9288.
Blanchard, A., Ohanian, V., and Critchley, D. (1989). The structure
and function of ?-actinin. J . Muscle Res. Cell Motil. 10, 280±289.
Structure of ?-Actinin Rod
Bru Ènger, A.T. (1992). X-PLOR, a System for Crystallography and
NMR, Version 3.1, Manual Edition. (New Haven, CT: Yale University
Bru Ènger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P.,
Grosse-Kunstleve, R.W., J iang, J .-S., Kuszewski, J ., Nilges, M., and
Pannu, N.S., et al. (1998). Crystallography and NMR system (CNS):
a new software suite for macromolecular structure determination.
Acta Crystallogr. D 54, 905±921.
Burridge, K., and Chrzanowska-Wodnicka, M. (1996). Focal adhe-
sions, contractility, and signaling. Annu. Rev. Cell Dev. Biol. 12,
Carpen, O., Pallai, P., Staunton, D.E., and Springer, T.A. (1992).
containing cytoskeleton and ?-actinin. J . Cell Biol. 118, 1223±1234.
Carugo, O., and Bordo, D. (1999). How many water molecules can
be detected by protein crystallography?. Acta Crystallogr. D 55,
Castresana, J ., and Saraste, M. (1995). Does Vav bind to F-actin
through a CH domain? FEBS Lett. 374, 149±151.
CCP4 (Collaborative Computing Project No. 4). (1994). The CCP4
suite: programs for protein crystallography. Acta Crystallogr. D 50,
Crawford,A.W.,Michelsen, J .W.,andBeckerle,M.C.(1992).Aninter-
action between zyxin and ?-actinin. J . Cell Biol. 116, 1381±1393.
Davison, M.D., and Critchley, D.R. (1988). ?-Actinins and the DMD
protein contain spectrin-like repeats. Cell 52, 159±160.
dela Fortelle,E., andBricogne, G.(1997).Maximum-likelihoodheavy
atom parameter refinement for multiple isomorphous replacement
and multiwavelength anomalous diffraction methods. Methods En-
zymol. 276, 472±494.
Diederichs, K. (1995). Structural superposition of proteins with un-
known alignment and detection of topological similarity using a six-
dimensional search algorithm. Proteins 23, 187±195.
Flood, G., Kahana, E., Gilmore, A.P., Rowe, A.J ., Gratzer, W.B., and
Critchley, D.R. (1995). Association of structural repeats in the
?-actinin rod domain. J . Mol. Biol. 252, 227±234.
Flood, G., Rowe, A.J ., Critchley, D.R., and Gratzer, W.B. (1997).
Further analysis of the role of spectrin repeat motifs in ?-actinin
dimer formation. Eur. Biophys. J . 25, 431±435.
Furey, W.B., and Swaminathan, S. (1997). PHASES-95: a program
package for the processing and analysis of diffraction data from
macromolecules. In Methods in Enzymology, Vol. 277, C.W. Carter
and R.M. Sweet, eds. (Orlando, FL: Academic Press), pp. 590±620.
Gilmore, A.P., Parr, T., Patel, B., Gratzer, W.B., and Critchley, D.R.
(1994). Analysis of the phasing of four spectrin-like repeats in
?-actinin. Eur. J . Biochem. 225, 235±242.
Gregorio, C.C., Granzier, H., Sorimachi, H., and Labeit, S. (1999).
Muscle assembly: a titanic achievement? Curr. Opin. Cell Biol. 11,
Grum, V.L., Li, D., MacDonald, R.I., and Mondrago Ân, A. (1999). Struc-
tures of two repeats of spectrin suggest models for flexibility. Cell
98, this issue, 523±535.
Heiska, L., Kantor, C., Parr, T., Critchley, D.R., Vilja, P., Gahmberg,
C.G., and Carpen, O. (1996). Binding of the cytoplasmic domain of
intercellular adhesion molecule-2 (ICAM-2) to ?-actinin. J . Biol.
Chem. 271, 26214±26219.
Higgins, D.G., Bleasby, A.J ., and Fuchs, R. (1992). CLUSTAL V:
improved software for multiple sequence alignment. Comput. Appl.
Biosci. 8, 189±191.
Horwitz, A., Duggan, K., Buck, C., Beckerle, M.C., and Burridge, K.
(1986). Interaction of plasma membrane fibronectin receptor with
talinÐ a transmembrane linkage. Nature 320, 531±533.
Imamura, M., Endo, T., Kuroda, M., Tanaka, T., and Masaki, T.(1988).
Substructure and higher structure of chicken smooth muscle
?-actinin molecule. J . Biol. Chem. 263, 7800±7805.
J ones, S., and Thornton, J .M. (1996). Principles of protein-protein
interactions. Proc. Natl. Acad. Sci. USA 93, 13±20.
J ones, T.A., Zou, J .Y., Cowen, S.W., and Kjeldgaard, M. (1991).
Improved methods for building protein models in electron density
maps and the location of errors in these models. Acta Crystallogr.
A 47, 110±119.
Kahana, E., and Gratzer, W.B. (1991). Properties of the spectrin-like
structural element of smooth-muscle ?-actinin. Cell Motil. Cytoskel-
eton 20, 242±248.
Kohn, W.D., Kay, C.M., and Hodges, R.S. (1998). Orientation, posi-
tional, additivity, and oligomerization-state effects of interhelical ion
pairs in ?-helical coiled-coils. J . Mol. Biol. 283, 993±1012.
Kraulis, P.J . (1991). Molscript: a program to produce both detailed
and schematic plots of protein structures. J . Appl. Crystallogr. 24,
Kuroda, M., Kohira, Y., and Sasaki, M. (1994). Conformational
change of skeletal muscle ?-actinin induced by salt. Biochim. Bio-
phys. Acta. 1205, 97±104.
Landon, F., Gache, Y., Touitou, H., and Olomucki, A. (1985). Proper-
ties of two isoforms of human blood platelet ?-actinin. Eur. J . Bio-
chem. 153, 231±237.
Laskowski, R.A., MacArthur, M.W., Moss, D.S., and Thornton, J .M.
(1993). PROCHECK: a program to check the stereochemical quality
of protein structures. J . Appl. Crystallogr. 26, 283±291.
Lyu, P.C., Gans, P.J ., and Kallenbach, N.R. (1992). Energetic contri-
bution of solvent-exposed ion pairs to ?-helix structure. J . Mol. Biol.
McLachlan, A.D., and Stewart, M. (1975). Tropomyosin coiled-coil
interactions: evidence foran unstaggered structure. J . Mol. Biol. 98,
Merritt, E.A., and Murphy, M.E.P. (1994). Raster3D Version 2.0: a
program for photorealistic molecular graphics. Acta Crystallogr. D
Mukai, H., Toshimori, M., Shibata, H., Takanaga, H., Kitagawa, M.,
Miyahara, M., Shimakawa, M., and Ono, Y. (1997). Interaction of
PKN with ?-actinin. J . Biol. Chem. 272, 4740±4746.
Nicholls, A., Sharp, K.A., and Honig, B. (1991). Protein folding and
association: insights from the interfacial and thermodynamic prop-
erties of hydrocarbons. Proteins 11, 281±296.
Ohtsuka, H., Yajami, H., Maruyama, K., and Kimura, S. (1997). The
N-terminal Z repeat 5 of connectin/titin binds to the C-terminal re-
gion of ?-actinin. Biochem. Biophys. Res. Commun. 235, 1±3.
Otey, C.A., Pavalko, F.M., and Burridge, K. (1990). An interaction
between ?-actinin and the ?1 integrin subunit in vitro. J . Cell Biol.
Otwinowski, Z., and Minor, W. (1997). Processing ofX-ray diffraction
data collected in oscillation mode. Methods Enzymol. 276, 307±326.
Pannu, N.S., and Read, R.J . (1996). Improved structure refinement
through maximum likelihood. Acta Crystallogr. A 52, 659±668.
Parks, T.D., Leuther, K.K., Howard, E.D., J ohnston, S.A., and Dou-
gherty, W.G. (1994). Release of proteins and peptides from fusion
proteins using a recombinant plant virus proteinase. Anal. Biochem.
Pascual, J ., Castresana, J ., and Saraste, M. (1997a). Evolution of
the spectrin repeat. Bioessays 19, 811±817.
Pascual, J ., Pfuhl, M., Walther, D., Saraste, M., and Nilges, M.
(1997b). Solutionstructure ofthe spectrinrepeat: a left-handed anti-
parallel triple-helical coiled-coil. J . Mol. Biol. 273, 740±751.
Pavalko, F.M., Walker, D.M., Graham, L., Goheen, M., Doerschuk,
C.M., and Kansas, G.S. (1995).The cytoplasmic domainofL-selectin
interacts with cytoskeletal proteins via ?-actinin: receptor position-
ing in microvilli does not require interaction with ?-actinin. J . Cell
Biol. 129, 1155±1164.
Read, R.J . (1986). Improved Fourier coefficients for maps using
phases from partial structures with errors. Acta Crystallogr. A 42,
Rief, M., Pascual, J ., Saraste, M., and Gaub, H.E. (1999). Single
molecule force spectroscopy of spectrin repeats: low unfolding
forces in helix bundles. J . Mol. Biol. 286, 553±561.
SÏali, A., and Blundell, T.L. (1993). Comparative protein modelling by
satisfaction of spatial restraints. J . Mol. Biol. 234, 779±815.
Speicher, D.W., and Marchesi, V.T. (1984). Erythrocyte spectrin is
comprised ofmany homologous triple helicalsegments. Nature 311,
Cell Download full-text
Speicher, D.W., Weglarz, L., and DeSilva, T.M. (1992). Properties of
human red cell spectrin heterodimer (side-to-side) assembly and
identification of an essential nucleation site. J . Biol. Chem. 267,
Taylor, K.A., and Taylor, D.W. (1993). Projection image of smooth
muscle?-actininfromtwo-dimensionalcrystals formed onpositively
charged lipid layers. J . Mol. Biol. 230, 196±205.
Trave, G., Pastore, A., Hyvonen, M., and Saraste, M. (1995). The
C-terminaldomainof?-spectrinis structurally related to calmodulin.
Eur. J . Biochem. 227, 35±42.
Viel, A., and Branton, D. (1994). Interchain binding at the tail end of
the Drosophila spectrin molecule. Proc. Natl. Acad. Sci. USA 91,
Wachsstock, D.H., Wilkins, J .A., and Lin, S. (1987). Specific interac-
tion of vinculin with ?-actinin. Biochem. Biophys. Res. Commun.
Winder, S.J ., Gibson, T.J ., and Kendrick-J ones, J . (1995). Dystropin
and utrophin: the missing links! FEBS Lett. 369, 27±33.
Winkler, J ., Lu Ènsdorf, H., and J ockusch, B.M. (1997). Flexibility and
fine structure of smooth-muscle ?-actinin. Eur. J . Biochem. 248,
Wyszynski, M., Lin, J ., Rao, A., Nigh, E., Beggs, A.H., Craig, A.M., and
Sheng, M.(1997).Competitive binding of?-actininand calmodulinto
the NMDA receptor. Nature 385, 439±442.
Xia, H., Winokur, S.T., Kuo, W.L., Altherr, M.R., and Bredt, D.S.
(1997). Actinin-associated LIM protein: identification of a domain
interactionbetweenPDZ and spectrin-like repeatmotifs. J . CellBiol.
Yan,Y., Winograd,E.,Viel, A.,Cronin,T., Harrison,S.C., andBarnton,
D. (1993). Crystal structure of the repetitive segments of spectrin.
Science 262, 2027±2030.
Young, P., Ferguson, C., Banuelos, S., and Gautel, M. (1998). Molec-
ularstructure ofthe sarcomeric Z-disk: two types oftitininteractions
lead toanasymmetricalsorting of?-actinin.EMBOJ .17,1614±1624.
Protein Data Bank ID Code
The R2R3 structure reported in this paper has been deposited in
the Protein Data Bank under ID code 1QUU.