Proc. Natl. Acad. Sci. USA Vol. 74, No. 5, pp. 2021-2025, May 1977 Cell Biology

Interaction of filamin with F- in solution (contractile /cytoskeletal components/ organization) KUAN WANG* AND S. J. SINGER Department of Biology, University of California at San Diego, La Jolla, California 92093 Contributed by S. J. Singer, February 22, 1977

ABSTRACT Filamin is a major high-molecular-weight in the gel and electrode buffers. Coomassie brilliant blue in smooth muscle which was recently identified and staining intensities at 550 nm were measured by a Gilford isolated [Wang, K., Ash, J. F. & Singer, S. J. (1975) Proc. Natl. spectrophotometer equipped with a gel scanner, and were Acad. Sci. U.S.A. 72,4483-4486]. In the present studies, we show that highly purified chicken gizzard filamin and muscle F-actin converted to relative protein content by calibration electro- react in solution to form aggregates containing both proteins. phoresis experiments carried out with known mixtures of the Occasionally, these aggregates coagulate and contract into a filamin and actin proteins. dense gel in the absence of MgATP or CaATP. Immunofluo- The filamin-actin aggregate shown in Fig. 3 was examined rescence and electron microscopic studies suggest that the F- with a Zeiss Photomicroscope III equipped with Nomarski actin filaments are collected into fiber bundles and a crosslinked fiber meshwork by the binding of filamin molecules. These optics and an epifluorescence attachment (III RS) for obser- studies suggest that the function of filamin in intact cells may vations of the immunofluorescent staining of filamin in the be to regulate the ultrastructural state of F-actin filaments in aggregate by the indirect method. Details of the specimen a variety of dynamic cellular processes. preparation are given in the legends for Fig. 3. The filamin- actin aggregate, and the two proteins before mixing, were ex- We have recently reported the detection, isolation, and char- amined by electron microscopy after negative staining, as de- acterization of a new high-molecular-weight protein in smooth scribed in the legend of Fig. 4. A Philips model 300 electron muscle and in a variety of nonmuscle cells, which we have microscope at 60 kV was used. named filamin (1, 2). Its name derives from the finding (1), by specific immunofluorescence techniques, that filamin is asso- RESULTS ciated with intracellular filamentous structures which also contain and actin. The apparent ubiquity of this protein, Aggregation of F-Actin by Filamin. When clear solutions and its association with filaments, suggests some important of highly purified filamin and actin, at concentrations of a few intracellular mechanochemical role for filamin, but its function tenths mg/ml in buffer B, were mixed at appropriate ratios at is not yet known. Although filamin is associated with filaments room temperature, gel-like aggregates formed immediately. in vivo, isolated pure filamin does not appear itself to form Usually, the aggregates formed discrete gel particles which filaments in vitro (2). We therefore suggested the possibility remained suspended in the solution. But occasionally, (two times (1) that filamin may interact and combine with filamentous in 10 preparations) the entire aggregate started to coagulate and structures formed by other components, in particular, F-actin. slowly contracted into a dense translucent gel. This phenome- The experiments reported in this paper provide evidence for non is shown in Fig. 1. Why most of the aggregated samples such an interaction in vitro and suggest possible physiological remained suspended and others coagulated is not clear. No functions for filamin. significant differences were observed with the two different that were prepared. MATERIALS AND METHODS If the aggregates that formed remained suspended, then within 10 min after formation they could be sedimented by Filamin from fresh chicken gizzard, the IgG fraction of rabbit low-speed centrifugation in an SS-34 rotor of a Sorvall RC2-B antifilamin antibodies, and the fluorescein-conjugated goat centrifuge operated at 12,000 rpm for 15 min. These conditions antibodies to rabbit IgG were prepared as described elsewhere freed the supernatant of aggregates, as judged by light mi- (refs. 1 and 2; K. Wang and S. J. Singer, unpublished data). croscopy, without sedimenting significant amounts of protein Actins from chicken breast muscle and rabbit back muscle were from solutions containing either filamin or F-actin alone. Under purified according to Spudich and Watt (3). these conditions, about 10-30% of the initial protein in these Most of the studies reported here were carried out in a buffer filamin-actin mixtures was sedimented. The sediment and the containing 0.1 M KC1/0.02 M Tris-HCl/0.2 mM dithiothreitol, supernatant, when analyzed by gel electrophoresis, both con- pH 7.4 (buffer B). Proteins were dialyzed extensively against tained the two proteins and no others. this buffer before analysis. When indicated, the ionic strength In the cases where the aggregates coagulated, much larger was adjusted by adding an appropriate amount of 3.0 M KC1 amounts of protein (70% for the sample shown in Fig. 1) were in buffer B. Protein concentrations were measured by the sedimented by the low centrifugal force field described above. method of Lowry et al. (4) as modified by Hartree (5). Poly- The gel electrophoresis patterns of the sediment and superna- acrylamide gel electrophoresis in the presence of sodium do- tant fractions of the mixture of Fig. 1 are shown in Fig. 2. The decyl sulfate (NaDodSO4) was performed according to Fair- coagulated aggregates (Fig. 2D) had no detectable amount of banks et al. (6) on 4% acrylamide gels using 0.2% NaDodSO4 myosin, a-, or any protein bands other than filamin and actin. Abbreviations: buffer B, 0.1 M KC1, 0.02 M Tris-HCl, 0.2 mM di- thiothreitol, pH 7.4; NaDodSO4, sodium dodecyl sulfate. The filamin-actin aggregation process was studied quanti- * Present address: Department of Chemistry and Clayton Foundation tatively as a function of the initial ratio of the two proteins. The Biochemical Institute, University of Texas at Austin, Austin, TX results are recorded in Table 1. As the initial weight ratio of 78712. filamin to actin was increased up to about 1.1-1.2, the weight 2021 Downloaded by guest on September 25, 2021 2022 Cell Biology: Wang and Singer Proc. Nati. Acad. Sci. USA 74 (1977)

A B C D

FIG. 1. Coagulation of the aggregate formed upon mixing filamin and actin. Filamin and actin in buffer B were mixed at room temperature (final protein concentration: filamin, 100,gg/ml; actin, 400 jg/ml) in a polycarbonate tube. The coagulated gel was detected by observing the air bubbles trapped in the gel (indicated by the arrow). Photographs were taken at 5 min (tube A), 20 min (tube B), 40 min (tube C), and 60 min (tube D) after mixing.

ratio in the sedimented aggregates increased correspondingly; sedimentable in the low centrifugal force fields increased with but, above this value, the ratio in the sediment remained increasing total protein concentration in the initial mixture practically constant or increased slightly in the range of 1.1-1.3. (from 3% at 0.1 mg/ml to 27% at 2.0 mg/ml). At a given total In different experiments, this value fell in the range 1.0-1.5. protein concentration, the amount of sedimentable aggregate These data suggest that a nearly stoichiometric binding of decreased as the KCl concentration was increased; above 0.5 filamin to actin is involved in the aggregation phenomenon, M KC1 in buffer B no aggregate was detectable. On the other with a mole ratio of 1 filament molecule (molecular weight hand, once formed at lower salt concentrations, the aggregates 500,000) to 8 to 12 actin monomer molecules (molecular weight were stable in 0.6 M KCI for a prolonged period, but dissolved 42,000). immediately in 0.6 M KI, presumably due to the depolymer- In separate experiments, there was no detectable interaction ization of F-actin to G-actin (see below). Low temperature (40) found when purified filamin and smooth muscle myosin of had no appreciable effect on the amount of aggregate, nor did chicken gizzard were mixed under similar conditions. Mg2 , Ca2 , MgATP, CaATP, or EDTA, up to 5 mM concen- Factors Affecting the Extent of Aggregation. A series of trations. Cytochalasin B up to 25 jug/ml had no effect. Heating experiments was carried out at a fixed filamin-to-actin weight both protein solutions at 1000 for 5 min abolished their aggre- ratio of 1.0 in buffer B, in which other variables were altered. gation properties on mixing. It was found that the percent of the total protein that was The physical state of the actin was examined as a factor in

A B C D Table 1. Interaction of filamin with actin* Weight ratio of filamin to actin i'- *-F Filamin Original mixture Supernatant Sedimentt 0.15 0.13 0.20 0.30 0.25 0.28 0.50 0.58 0.45 0.75 0.90 0.80 1.0 1.1 0.90 1.25 1.4 1.1 1.75 1.6 1.15 2.50 2.7 1.25 3.0 2.8 1.37 3.5 3.6 1.30 Si_ Actin * Results of a typical experiment. Various amounts of filamin were added with rapid mixing to a fixed amount (1 mg/ml) of chicken skeletal muscle actin in buffer B to yield the same final volume. The samples were allowed to stand at room temperature for 45 min be- fore they were centrifuged at 12,000 rpm for 15 min in a SS-34 rotor. The weight ratios of filamin to actin in the supernatant and sedi- ment were determined by quantitative NaDodSO4/gel electropho- FIG. 2. NaDodSO4/polyacrylamide gel electrophoresis of actin resis. (A), filamin (B), the supernatant of the sample shown in Fig. 1 (C), t These values have been corrected by subtracting blank values de- and the aggregate of the sample shown in Fig. 1 (D). Electrophoretic termined in parallel experiments with either filamin or actin gels were overloaded to show the absence of myosin in the sample. omitted. Blank corrections were usually approximately 10%. Downloaded by guest on September 25, 2021 Cell Biology: Wang and Singer Proc. Natl. Acad. Sci. USA 74 (1977) 2023 the aggregation process. G-actin in buffers containing 5 mM Tris-HCI/0.5 mM 2-mercaptoethanol, with and without 0.2 mM CaATP at pH 7.5-8.5, was mixed with filamin dialyzed against the same buffer. No F-actin or aggregates of actin fibers could be detected by light and electron microscopy in such mixtures. Subsequently raising the KC1 concentration to 0.1 M immediately induced the formation of aggregate. The results suggest that filamin interacts with and aggregates F-actin, but does not induce the polymerization of, or aggregation with, G-actin. Structure of the Aggregates. Although the near stoichiom- etry of the filamin-actin compositionof theaggregates (Table 1) suggested that a direct interaction of the two proteins was involved, it was possible that the results could be explained by a coprecipitation of separate actin and filamin structures. Mi- croscopic examination of the aggregates was therefore under- taken. Under all conditions tested, the aggregate under light microscopy contained fiber bundles of various lengths and thicknesses (Fig. 3A). After staining the filamin in the aggregate by the indirect immunofluorescence technique, as described in the legend of Fig. 3, it became evident that in addition to the fiber bundles there were sheet-like connections between them with both the bundles and the sheets uniformly fluorescent stained (Fig. 3B). If the antifilamin antibody was preabsorbed FIG. 3. Localization of filamin in the aggregate by fluorescent with filamin and then in antibody staining. An aggregate similar to that shown in Fig. 1 was used the immunofluorescent staining adsorbed onto a polyornithine-coated cover glass and then stained of the aggregates, no significant fluorescence was observed. with antifilamin antibodies and indirect immunofluorescence as de- At the electron microscopic level of resolution, preparations scribed in the text. (A) Nomarski micrograph. (B) Fluorescence mi- of aggregates subjected to negative staining, as described in the crograph. (X170.) legend of Fig. 4, revealed dense bundles of fibers connected by sheets of extended fibrous meshworks At (Fig. 4C). higher On the other we do not magnification (Fig. 4 D and E) the fibrous meshwork was seen hand, yet understand the details of to consist of thin fibers of the diameter of F-actin fila- this mechanism. Why the filamin-actin aggregation only oc- single led to and contraction ments which were arranged in a roughly orthogonal matrix casionally gross coagulation is not known. More-or-less of the size of filamin The denaturation of the two proteins, a possible requirement (Fig. 4C). spherical particles for an unknown molecules 200 A were attached to low-molecular-weight cofactor, and the mul- (about diameter) frequently forms of actin are the factors these filaments. In favorable areas, these attached particles tiple (10-12) among which might to be in *have been responsible for the complexity of the results. appeared periodic array (Fig. 4D). By contrast, at the Despite these uncertainties, certain characteristics same protein concentration, F-actin alone consisted of single however, randomly oriented fiber without attached particles (Fig. 4A), of the filamin-actin interaction have been determined. The but never as bundles of fibers; filamin alone consisted of interaction of the two proteins is at the molecular level and more-or-less spherical particles of different extents of aggre- involves F-actin, not G-actin. The aggregates are not simply gation, but with no detectable fibrous structure (Fig. 4B). coprecipitates of two independently aggregating proteins, These morphological observations support the view that the because the immunofluorescence studies show (Fig. 3) that fil- are the result of a direct molecular interaction amin is uniformly dispersed over the fibrous bundles and aggregates of meshwork filamin and F-actin, and not simply due to an agglomeration of the aggregate. The same inference can be drawn of separate precipitates of the two proteins. from the electron micrographs of the aggregates (Fig. 4). Furthermore, filamin does not simply catalyze the aggregation DISCUSSION of F-actin, because appreciable amounts of the filamnin were bound in the aggregate along with the actin. We have shown that purified filamin from chicken gizzard When the stoichiometry of binding of the two proteins was interacts in solution with skeletal muscle F-actin to form gel-like investigated, the mixture did not show the characteristics of a aggregates at physiological ionic strength. On occasion, a gross simple two-component system. If it had, then in mixtures coagulation of these aggregates occurred, followed by a slow containing excess filamin, all of the F-actin should have been contraction of the coagulated gel (Fig. 1). Such a gel contraction bound in aggregates. Instead, the F-actin in the aggregates is generally characteristic of force-generating macromolecular appeared to become saturated with filamin while most of the systems, such as the actomyosin system of striated muscle (7) F-actin remained in the supernatants (Table 1). The reasons for or the cytoplasmic contractile system in acanthamoeba (8) and this complex binding behavior are not known; the same factors macrophages (9). In the acanthamoeba and macrophage sys- cited above in connection with the irreproducible coagulation tems, however, myosin is present and the gel contraction is process might be involved in these binding studies as well. ATP-dependent, as would be expected of an actomyosin-like Nevertheless, the relatively constant composition of the ag- contractile process. On the contrary, the contractile process gregates at high ratios of filamin to actin (Table 1) suggests that exhibited in Fig. 1 involved only the proteins filamin and actin a specific binding of the two proteins is involved which is sat- (Fig. 2D) and was not affected by ATP. These results therefore urated when 1 dimeric filamin molecule is bound per 8-12 suggest that a new type of force-generating mechanism may monomer units of the F-actin in the aggregate. This is not too be involved in this system, different from the actomyosin different from the binding stoichiometry of some other con- type. tractile proteins to F-actin: , known to bind in the Downloaded by guest on September 25, 2021 2024 Cell Biology: Wang and Singer Proc. Natl. Acad. Sci. USA 74 (1977)

FIG. 4. Negatively stained electron micrographs of F-actin (A), filamin (B), the filamin-actin aggregate at low magnification (C), a portion of (C) at higher magnification (D), and a representative view from another sample (E). For (A), (B), (D), and (E), bars indicate 0.1 Mm; for (C), bar indicates 1 Am. In (C) densely stained bundles of actin fibers are seen near the left of the micrograph. The meshwork appears to consist of nearly parallel arrays of actin fibers running in two directions, as indicated by the white arrows. In (D), some actin fibers are seen decorated by globular material of molecular sizes similar to filamin. The small black arrowheads (v) point to the globular material and the blank arrow (v) indicates the bare portion ofthe same F-actin fiber. In (E)the decoration is so dense that only occasionally bare, thin fibers can be detected entering the aggregate (blank arrow). groove of the F-actin helix, has a binding ratio of 1 molecule molecules bind directly to specific sites on F-actin, and that at to 7 actin monomers (13), while a-actinin, which causes F-actin some of these sites, the filamin molecules crosslink the actin to aggregate into bundles at 00 (but not at 370) (14), binds in strands into either fiber bundles or loose meshworks. This the ratio of 1 molecule to 9-11 actin monomers (15). crosslinking process leads to the formation of the gel aggregate. The electron micrographs show that the F-actin filaments What happens, however, when the gel aggregates coagulate and in the filamin-actin aggregates are crosslinked into filament then contract, is not clear. It might be pointed out that filamin bundles and into loose meshworks of overlapping individual in solution exists as a dimer of apparently identical chains (2), F-actin filaments. More-or-less spherical knobs, presumably so that in principle an individual dimeric unit could bind si- consisting of individual filamin molecules, can be seen (Fig. 4D) multaneously to two actin strands and thereby crosslink along the F-actin strands, which in favorable views are in an them. apparently periodic array. Having demonstrated and partially characterized the in vitro All of the binding and morphological data presented are interaction of filamin and actin, we may inquire whether this therefore consistent with the proposal that, in vitro, filamin interaction has any function inside the living cell. The absence Downloaded by guest on September 25, 2021 Cell Biology: Wang and Singer Proc. Natl. Acad. Sci. USA 74 (1977) 2025 of filamin from skeletal muscle cells indicates that it is itin- Afte-rith& present studies were completed, a paper appeared volved in the actomyosin sliding filament mechanism nor is by Shizuta et al. (22) on the isolation and characterization of filamin essential for its control. On the other hand, the presence chicken gizzard filamin. Based on the findings reported in our of filamin in smooth muscle cells and in a wide range of non- initial publication (1), these authors proceeded to carry out muscle cells so far examined suggests that it does have an im- similar studies to those we had completed or had in progress (2). portant general role in cell physiology. It is of considerable in- They presented a single analytical ultracentrifuge experiment terest that studies of the intracellular localization of filamin by depicting a pellet sedimented from a solution containing filamin immunofluorescent techniques (ref. 1; M. Heggeness, K. Wang, and F-actin as evidence for an interaction of the two pro- and S. J. Singer, unpublished data) have shown the filamin to teins. be distributed in two forms: in long thick filaments and in a diffuse intracellular matrix. These two forms in vivo appear We thank Mrs. Donna Luong for her excellent technical assistance. to correspond closely to the fiber bundles and loose meshworks, We are grateful to Dr. T. P. Stossel for sending us a copy of ref. 9 before respectively, which we have described above in filamin-actin its publication. S.J.S. is an American Cancer Society Research Professor, aggregates in vitro. This correspondence encourages some and K.W. was a U.S. Public Health Service Postdoctoral Fellow speculation about the possible role of filamin in vivo. 1974-1976. Although myosin-like molecules are as ubiquitous as actin The costs of publication of this article were defrayed in part by the in eukaryotic nonmuscle cells, there are specialized regions payment of page charges from funds made available to support the within these cells that are apparently devoid of the myosin and research which is the subject of the article. This article must therefore yet can exhibit certain mechanochemical activities. For ex- be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 ample, in experiments to be published (M. Heggeness, K. Wang, solely to indicate this fact. and S. J. Singer, unpublished data), by using a double fluores- cence technique to observe the distribution of two macromol- ecules within the same cell, it has been shown that within the 1. Wang, K., Ash, J. F. & Singer, S. J. (1975) Proc. Natl. Acad. Sci. ruffles at the periphery of isolated cultured cells, and in the USA 72,4483-4486. 2. Wang, K. (1977) , in press. regions of cell-cell contact, actin and filamin are present, but 3. Spudich, J. A. & Watt, S. (1971) J. Biol. Chem. 246, 4866- myosin is not detectable. is also absent. It is well known 4871. that ruffles are highly motile portions of the cell periphery, and 4. Lowry, D. H., Rosebrough, N. J., Farr, A. L. & Randall, R. J. that immediately after the ruffles of two cells come in contact, (1951) J. Biol. Chem. 193, 265-275. this motility ceases (16). It has been shown (17, 18) that ac- 5. Hartree, E. F. (1972) Anal. Biochem. 48,422-427. companying this process there occurs a rapid reorganization 6. Fairbanks, G., Steck, T. L. & Wallach, D. F. H. (1971) Bio- of what appears to be a loose meshwork of individual F-actin chemistry 10, 2606-2617. filaments into arrays of filament bundles. Without specifying 7. Szent-Gyorgyi, A. (1947) Chemistry of Muscular Contraction the detailed mechanisms involved, we suggest that the inter- (Academic Press, New York). action 8. Pollard, T. D. (1976) J. Cell Biol. 68,579-601. of filamin with F-actin is critically involved in these 9. Stossel, T. P. & Hartwig, J. H. (1976) J. Cell Biol. 68, 602-619. filament reorganizations and motility changes in cell ruffles. 10. Gruenstein, E. & Rich, A. (1975) Biochem. Biophys. Res. Com- In addition, filamin-actin interactions may be more generally mun. 64, 472-477. involved in nonmuscle cell mechanochemistry, whenever dy- 11. Whalen, R. G., Butler-Browne, G. S. & Gros, F. (1976) Proc. Natl. namic and transitory actin-containing structures form and Acad. Scd. USA 73,2018-2022. dissociate subject to the needs of the cell. In order to test these 12. Rubinstein, P. A. & Spudich, J. A. (1977) Proc. Natl. Acad. Sci. speculations, more information is required about filamin-actin USA 74, 120-123. interactions in vitro, and in particular, what physiologically 13. Drabikowski, W., Nowak, E., Barylko, B. & Dabrowska, R. (1972) relevant factors control these interactions. Cold Spring Harbor Symp. Quant. Biol. 37,245-249. The relationship of filamin to other proteins that are reported 14. Kawamura, M., Masaki, T., Nonomura, J. & Maruyama, K. (1970) J. Biochem. 68,577-580. to interact with actin remains to be clarified. The actin-binding 15. Goll, D. E., Suzuki, A., Temple, J. & Holmes, G. R. (1972) J. Mol. protein isolated from rabbit pulmonary macrophages (9, 19) Biol. 67, 469-488. shows many similarities to chicken gizzard filamin, including 16. Abercrombie, M. (1961) Exp. Cell Res. Suppl. 8, 188-198. apparently the capacity to form gels with F-actin filaments, 17. Heaysman, J. E. M. & Pegrum, S. M. (1973) Exp. Cell Res. 78, although it was not demonstrated that gel contraction could 71-78. occur in the complete absence of myosin, as in Fig. 1. If filamin 18. Goldman, R. D., Schloss, J. A. & Starger, J. M. (1976) in Cell and the macrophage actin-binding protein are indeed the same, Motility, eds. Goldman, R. D., Pollard, T. & Rosenbaum, J. (Cold then: (i) we have demonstrated its presence in a wide range of Spring Harbor Laboratory, Cold Spring Harbor, NY), Vol. A, nonmuscle cells as well as in smooth muscle (1); and (ii) smooth pp. 217-245. 19. Stossel, T. P. & Hartwig, J. H. (1975) J. Biol. Chem. 250, muscle is a better source for large amounts of the protein than 5706-5712. are macrophages. Another protein that is reported to affect the 20. Tilney, L. G. & Detmers, P. (1975) J. Cell Biol. 66,508-520. properties of actin in vitro is erythrocyte (20, 21). Al- 21. Pinder, J. C., Bray, D. & Gratzer, W. B (1975) Nature 258, though spectrin is distinctly different from filamin, the two 765-766. proteins do share some interesting physical characteristics (2), 22. Shizuta, Y., Shizuta, H., Gallo, M., Davies, P., Pastan, I. & Lewis, but no definite relationship has as yet been established. M. S. (1976) J. Biol. Chem. 251, 6562-6567. Downloaded by guest on September 25, 2021