|Home | About | Journals | Submit | Contact Us | Français|
Filamin A (FLNa) is an actin-binding protein that cross-links F-actin into networks of orthogonally branched filaments. FLNa also directs the networks to integrins while responding to mechanochemical signaling pathways. Flexible, 160 nm long FLNa molecules are tail-to-tail dimers, each subunit of which contains an N-terminal calponin-homology/actin-binding domain connected by a series of 24 immunoglobulin (Ig) repeats to a dimerization site at their C-terminal end. Whereas the contribution of the CH-homology domains to F-actin affinity is weak (apparent Ka ~105), the binding of the intact protein to F-actin is strong (apparent Ka ~108), suggesting involvement of additional parts of the molecule in this association. Indeed, previous results indicate that Ig-repeats along FLNa contribute significantly to the strength of the actin filament interaction. In the current study, we used electron microscopy and three-dimensional reconstruction to elucidate the structural basis of the Ig-repeat–F-actin binding. We find that FLNa density is clearly delineated in reconstructions of F-actin complexed with either a 4 Ig-repeat segment of FLNa containing Ig-repeat 10 or with IgFLNa10 alone. The mass attributable to IgFLNa10 lies peripherally along the actin-helix over the N-terminus of actin subdomain 1. The interaction appears to be specific, since no other fragment of the FLNa molecule or individual Ig-repeats examined, besides ones with CH-domains, decorated F-actin filaments or were detected in reconstructions. We conclude that the combined interactions of CH-domains and the IgFLNa10 repeat provide the binding strength of the whole FLNa molecule and propose a model for the association of IgFLNa10 on actin filaments.
The cortical cytoskeleton regions of virtually all eukaryotic cells contain highly dynamic networks of actin filaments. In order to control cell shape and motility, the networks model and remodel, while responding to intracellular and/or extracellular chemical and mechanical signals.1–3 Typically signaling pathways lead to the binding or dissociation of members of a group of well over 100 distinct actin-binding proteins, which in turn specify the restructuring of the actin cytoskeleton.4 Some actin-binding proteins nucleate, sever or depolymerize filaments, thus regulating F-actin growth or disassembly. Others cross-link F-actin into tight bundles or looser networks, thus transforming cell superstructure and plasticity.1–4 Bivalent filamin A (FLNa), the first non-muscle actin-binding protein to be identified5 and the subject of our investigation, cross-links neighboring F-actin in cells into arrays of orthogonally oriented filaments6, while at the same time presenting surface domains for protein partner interactions that are primed to transform the resulting filament network further.7–11
The C-terminal ends of elongated FLNa subunits self-associate to form a flexible V-shaped homodimer. The tips of the resulting structure, at the N-termini of the dimerized chains, contain paired CH-domains (a.k.a. actin-binding domains or ABD), which are conserved modules found in many actin-binding proteins.12 They contribute, at least partially, to the actin-binding function of FLNa, and their presence at the two ends of the dimer can account for FLNa - F-actin cross-linking. 8,10,13
Each N-terminal CH-domain/ABD is separated from the C-terminal dimerization site by a ~80 nm long set of 24 Ig-segments arranged in tandem. Hinge domains separating Ig-repeats 15 and 16 as well as repeats 23 and 24 provide molecular flexibility. Pairing between repeats 16–17, 18–19, and 20–21 generate a globular C-terminal domain that can be distended by mechanical forces.14,15 Binding studies show that FLNa CH-domain constructs, like other members of the “spectrin ABD superfamily”, interact with F-actin with very low affinity (Kd ~ 1.7 × 10−5 M); however, the intact full-length FLNa dimer binds actin filaments with considerably higher avidity (Kd ~ 1.7 × 10−8 M).8,10 Hence, other regions of the molecule, presumably the FLNa Ig-repeats, are likely to contribute to the binding strength of intact FLNa for F-actin.8,10 In fact, Ig-repeats in muscle proteins such as paladin16,17, kettin18,19, myosin-binding protein C20–22, and titin23 all have been reported to interact with F-actin.
In our previous work, a library of FLNa fragments was generated and used to characterize filamin-actin binding by F-actin co-sedimentation.8 This work showed that constructs including both FLNa Ig-repeats 9–15 and the FLNa CH-domains displayed F-actin binding comparable to that of the full length protein, suggesting a role for Ig-repeat - actin interaction.8 However, the studies did not categorize the binding-site on FLNa further or locate the Ig-binding target on actin. In order to define filamin-actin association more completely, we have now examined potential binding interactions of Ig-domains structurally by electron microscopy and 3D reconstruction. Here F-actin filaments mixed either with multidomain FLNa fragments or with single Ig-repeats were studied. 3DEM of complexes containing Ig-repeat 10 (a mutational hot spot)24 showed distinct densities attributable to the Ig-domain which is bound over the N-terminus of actin sub-domain 1. FLNa constructs, besides those containing IgFLNa10 or CH-domains, did not show obvious signs of F-actin labeling. The general significance of our findings is discussed.
Long FLNa fragments containing 6 to 8 Ig-repeats, particularly those with N-terminal CH-domains, bundle F-actin and were not suitable for further analysis. Four relatively short FLNa constructs, which represent most of the molecule (viz. FLNa segments: IgFLNa8-11, IgFLNa12-15, IgFLNa16-23, and ABD-FLNaIg1-4), were chosen for study since they do not bundle F-actin. In addition, individually expressed Ig-repeats 9, 10 and 17 were examined (IgFLNa9, IgFLNa10, IgFLNa17 (nomenclature from ref. 8)). All constructs were mixed with 1 μM F-actin in 2–5:1 molar excess (molar ratio to actin subunits) to promote effective binding without causing significant protein background interference in electron microscope images. Electron micrographs of negatively stained samples confirmed that none of these constructs cross-link F-actin, and thus the filaments observed were well separated from each other as required for subsequent analysis. Direct inspection of the micrographs suggested an increased filament diameter brought about by the presence of, for example, fragments IgFLNa8-11, ABD-FLNaIg1-4, and IgFLNa10 (Fig. 1). However, no discrete structures derived from FLNa were obvious in any of the micrographs; hence, 3D image alignment and reconstruction was required to characterize potential binding interactions in the filaments.
Reconstructions generated from FLNa labeled F-actin reveal well-demarcated helically arranged actin-subunits and clearly defined actin subdomain structure (Fig. 2). Additional mass is seen on the surface of F-actin in reconstructions of mixtures containing IgFLNa10 and IgFLNa8-11 (Fig. 2b, g, arrows). These extra densities are present on the “top” of subdomain 1 of each actin subunit. The IgFLNa10 label forms a fairly compact density that caps subdomain 1, which partially obscures subdomain 2 (Fig. 2b, j, ,4a).4a). The mass formed by IgFLNa8-11, the 4 repeat fragment containing IgFLNa10, displays broader extra density on the upper aspect of actin subdomain 1 that extends across subdomain 1 in the direction of subdomain 3 (Fig. 2g, arrows). In contrast, F-actin decorated with constructs representing Ig-repeats 12–15 (Fig. 2h) or 16–23 (not shown) do not show signs of extra density on actin subunits, nor do single Ig-repeats 9 or 17 show structural evidence of binding (Fig. 2e, f). As expected, the mixed domain construct consisting of the FLNa CH-domain ABD and, in the case tested, the first four FLNa Ig-repeats (ABD-IgFLNa1-4) does indicate F-actin-labeling, evidently derived from the CH-domain interactions on F-actin (Fig. 2i, 2k), corroborating previous reports on the structural binding of FLNa CH-domain to actin.25 None of the constructs tested changed the average twist values of the actin filament (noted in Fig. 2), which is consistent with previous results indicating that phalloidin does not interfere with filamin-binding to F-actin26.
The densities contributing to the F-actin – IgFLNa10 reconstructions, including those derived from IgFLNa10, are statistically significant at 95 to 99 percent confidence levels. Moreover, IgFLNa10 contribution can also be identified as discrete “difference densities” when maps of control pure F-actin are subtracted from those of F-actin – IgFLNa10 filaments (Fig. 3a, b). These difference densities fall within outlines of the F-actin – IgFLNa10 reconstruction and are themselves statistically significant at greater than 99% confidence levels. In contrast, the IgFLNa10 difference densities fall outside the boundaries of F-actin control reconstructions (Fig. 3c, d).
Despite the relatively low number of filaments in half-data sets of F-actin – IgFLNa10, IgFLNa10 still is readily detected on actin subdomain 1 in reconstructions generated from such divided data (Fig. 2c, d, arrows). In fact, the IgFLNa10 density is a consistent feature of reconstructions of individual filaments of the data set. In addition, a small extra density derived from IgFLNa10 occasionally is detected on subdomains 3 and 4 on the outer border of actin (Fig. 2b, d, open arrows).
Docking atomic models of F-actin27 into the reconstructions of IgFLNa10 decorated filaments leaves considerable density over the N-terminus of F-actin unoccupied (Fig. 4). The remaining volume can be largely accommodated by the crystal structure of IgFLNa1028 which fits snugly into the remaining space (Fig. 4). However, the polarity and sidedness of the Ig-structure in the reconstruction is indeterminate, since neither the atomic model of IgFLNa10 nor the EM maps provide sufficient landmarks or asymmetries that can be used to match respective geometrical features with precision. In fact, the extra volume attributable to IgFLNa10 binding is about 30% broader than the dimensions of the crystal structure, suggesting azimuthal variance in the localization of the Ig-repeat on the surface of actin (Fig. 4c).
The N-terminus of actin contains a string of negatively charged amino acids. If electrostatic interactions with the actin N-terminus are a driver of IgFLNa10-actin binding, then only one interface of the IgFLNa10 structure contains a suitable binding surface of complementary basic residues, and includes Lys1162, Lys1164, Arg1172, Lys1234, Lys1246 (cf. pdb ID# 3rgh).28 Interestingly, these residues outline a groove on IgFLNa10 that may be the target for the actin’s acidic residues (Fig. 4). Indeed, the N-terminal residues on actin represent one of the least conserved regions of this otherwise stringently conserved molecule29, suggesting that actin isoform and Ig-repeat specificity may account for filamin sorting to one or another cell cytoskeletal domain. However, many of these basic amino acids are themselves conserved in the other Ig repeats including IgFLNa9, 17, 16-23, which do not interact with F-actin, suggesting that neighboring residues, in addition to these basic amino acids, play a role in mediating this interaction or in producing a preference for solvent. Higher resolution structures are necessary to resolve this interface in greater detail. Further studies are also necessary to determine how the disease-causing mutations affect the structure of IgFLNa10 and whether they alter its F-actin interaction.
The studies reported here directly reveal an Ig-domain – F-actin interaction structurally for the first time. Despite relatively low sequence similarity, members of the Ig-domain superfamily are structurally related by a common “Ig-fold” consisting of two sheets of anti-parallel β-strands that are sandwiched together.16 Given that Ig-domains have varied amino acid sequences, it is not surprising that only certain FLNa Ig-modules, in the case of our studies, Ig10FLNa, bind to F-actin with demonstrable affinity. Nevertheless, we cannot exclude the possibility that macromolecular crowding in a cellular environment may increase the effective concentration of other FLNa Ig-modules sufficiently to favor actin interactions that went undetected under our conditions in vitro.
Based on the current and previous work, we propose a model of FLNa cytoskeletal interaction in which neighboring but weakly interacting CH- and Ig-domains cooperatively promote FLNa binding to F-actin. This process then repeats itself as FLNa populates F-actin and cross-links adjoining filaments to form a mechanically cohesive network. However, because local FLNa-actin interactions are weak, chemical or mechanical signaling can easily perturb these networks during cytoskeletal remodeling. The general binding strategy discussed here, in which an elongated molecule utilizes multiple well-separated but weakly interacting domains to bind to F-actin, is not limited just to filamin-binding function but may also operate in myosin-binding protein C – thin filament association, where multiple Ig-repeats and the M-motif appear to reinforce each other during thin filament binding20,21,. Similarly, in the case of elongated dystrophin and utrophin, spectrin-repeats distal to CH-domains may enhance actin-binding.30–33 In fact, the well-established formula for tropomyosin-binding to thin filaments involves multiple weakly interacting pseudo-repeats linked to successive actin subunits along filaments.34,35 Thus actin network patterning based on tandem weakly interacting protein domains may be a commonly used strategy to style the fabric of the actin cytoskeleton.
This work was supported by National Institutes of Health Program Project Grant P01 HL86655 to Kathleen G. Morgan (W.L., principal investigator), R01 HL036153 to W.L., R01 HL089224 to J.H.H., and the Harvard University Science and Engineering Committee Seed Fund for Interdisciplinary Science to F.N.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.