Proc. Natl. Acad. Sci. USA Vol. 86, pp. 5773-5777, August 1989 polymerization induces a shape change in actin-containing vesicles (actin/actin-binding /lipid vesicles) JORGE DANIEL CORTESE*, BILL SCHWAB III, CARL FRIEDEN, AND ELLIOT L. ELSON Department of Biochemistry and Molecular Biophysics, Washington University School of Medicine, Saint Louis, MO 63110 Contributed by Carl Frieden, May 8, 1989

ABSTRACT We have encapsulated actin rilaments in the filament polymerization can drive lamellar extension during presence and absence of various actin-binding proteins into cellular locomotion (7). lipid vesicles. These vesicles are approximately the same size as animal cells and can be characterized by the same optical microscopic and mechanical techniques used to study cells. We MATERIALS AND METHODS demonstrate that the initially spherical vesicles can be forced Materials. Actin was prepared from rabbit skeletal muscle into asymmetric, irregular shapes by polymerization of the by the method of Spudich and Watts (8) and then gel-filtered actin that they contain. Deformation of the vesicles requires through Sephadex G-150 (9) and labeled with (iodoacetami- that the actin filaments be on average at least =0.5 ,um long as do)tetramethylrhodamine (IATR) (10). Plasma was shown by the effects of gelsolin, an actin rilament-nucleating purified from rabbit plasma (4, 11). Filamin was purified by . Filamin, a filament-crosslinking protein, caused the a method (12) modified as described (4). Fluorescein-labeled surfaces of the vesicles to have a smoother appearance. Het- dextran (Mr 36,500) was purchased from Sigma, dissolved at erogeneous distribution of actin filaments within the vesicles is 16 mg/ml in double-glass-distilled water, dialyzed against 1 caused by interfilament interactions and modulated by gelsolin mM EDTA overnight, and stored at 20'C. All reagents other and ifiamin. The vesicles provide a model system to study than actin and actin-binding proteins were obtained from control ofcell shape and cytoskeletal organization, membrane- Sigma. interactions, and cytomechanics. Incorporation ofActin and Actin-Binding Proteins into Lipid Vesicles. In a modification of a previously described method (13), loaded vesicles were prepared by injection of a diethyl The shapes and mechanical properties of animal cells are ether aqueous solution into a heated solution of the protein to governed mainly by systems of cytoplasmic filaments col- be encapsulated. To minimize the time for encapsulation and lectively termed the cytoskeleton (1, 2). Ofthese systems, the the extent of protein denaturation, we used both a lower actin system is the principal determinant of diethyl ether/water ratio, 1:10 (vol/vol), than traditional cellular viscoelastic properties (B.S. and E.L.E., in prepa- methods (13, 14) and a briefer (<5 min) exposure to 550C ration) and is most directly involved in driving mechanical temperatures. A mixture of 80% phosphatidylcholine (Ptd- processes such as locomotion, cytokinesis, and phagocytosis Cho) and 20% phosphatidylethanolamine (PtdEtn) was se- (2, 3). As cells perform these functions, the organization of lected to minimize problems oflipid-phase formation (15) and the actin cytoskeleton changes, probably under the control of maximize K+ and Cl- permeability (16). This mixture also actin-binding proteins that regulate the length and extent of avoided aggregation of vesicles (apparent with PtdCho or crosslinking of the filaments (3-5). A model system in which PtdEtn/phosphatidylserine vesicles) (17). The negatively cytoskeletal components could be reconstituted inside vesi- charged surface prevented actin polymerization directly from cles comparable in size to cells would be useful for studies of the membrane (18). PtdEtn and PtdCho at a molar ratio of 1:4 the regulation of cytoskeletal organization and the determi- together with 0.1 mg of valinomycin per mg of total phos- nation of cellular mechanical properties by the cytoskeleton. pholipid (0.35 mM final concentration) were dried down from We have developed a model system of this kind for the stock solutions immediately prior to use and dissolved into actin filament system. Actin and actin-binding proteins have -100 gl of water-washed diethyl ether. One milliliter of the been encapsulated in lipid vesicles large enough (up to 20 jim protein solution (actin with or without filamin or gelsolin in in diameter) to be characterized by optical microscopy. The 0.2 mM CaCl2/1.5 mM NaN3/0.2 mM ATP/2 mM Tris HC1, vesicles are large enough to be studied with biophysical pH 8.0; G buffer) was warmed to 52-550C in a water bath, and techniques that measure, for example, actin diffusion by the lipid stock solution was injected in small aliquots with fluorescence photobleaching recovery (FPR) (6) or mechan- gentle shaking (total time, <5 min). The lipid solution was ical properties (1) ofindividual vesicles. The reconstitution of injected at the bottom of the tubes containing the protein actin filaments with actin-binding proteins allows study ofthe mixtures. In all reconstitutions the actin concentration was 48 effects of the latter on filament organization and distribution 1LM. The vesicle-containing solution was cooled to room and the ability of the actin gel to drive morphological temperature, and the remaining ether was evaporated by changes. In this work we have investigated the effects of flushing the solution with nitrogen. gelsolin, which restricts the length of actin filaments, and of Polymerization of Actin Inside Lipid Vesicles. The KCI filamin, which crosslinks actin filaments (5). A striking ob- concentration of the vesicle-containing buffer was increased servation is that vesicles are deformed when the actin inside to 100 mM, and K+ was rapidly equilibrated across the them polymerizes. The extent ofthe deformation depends on membrane by valinomycin, thus polymerizing the encapsu- the lengths of the resulting filaments. This observation pro- lated actin. The high concentration of valinomycin ensured vides experimental evidence for the speculation that actin that only a few transport cycles (2-3) were needed to equil-

The publication costs of this article were defrayed in part by page charge Abbreviation: FPR, fluorescence photobleaching recovery. payment. This article must therefore be hereby marked "advertisement" *Present address: Department of Cell Biology and Anatomy, Uni- in accordance with 18 U.S.C. §1734 solely to indicate this fact. versity of North Carolina, Chapel Hill, NC 27599-7090. 5773 Downloaded by guest on September 26, 2021 5774 Biochemistry: Cortese et al. Proc. Natl. Acad. Sci. USA 86 (1989) ibrate K+. Valinomycin uniport avoids exchange of K+ for RESULTS other cations and the possible perturbation of the actin gel structure from resulting pH changes (19). Incorporation and Polymerization of Actin in Vesicles. A Vesicles were assayed for actin polymerization and mixture of 10% rhodamine-labeled/90% unlabeled actin, changes of vesicle shape after diluting 100 Al of the vesicle alone or with actin-binding proteins, was encapsulated by solution into 900 tul of either nonpolymerizing G buffer or injecting a diethyl ether solution of phospholipids and the polymerizing buffer (G buffer containing 100 mM KCl). This ionophore valinomycin into the warmed protein solution. A 10-fold dilution reduced the background fluorescence to a heterogeneous preparation with a high number of oligolamel- negligible level compared to vesicle fluorescence. The diluted lar vesicles with diameters between 2 and 20 ,m was ob- vesicle-containing solution was incubated at room tempera- tained. Fluorescence microscopy of the vesicles showed ture for 1 hr to allow actin polymerization. To eliminate actin encapsulation. When the potassium level within the background fluorescence in FPR experiments, we separated vesicles was raised by the addition of 100 KCI to the actin-containing vesicles from the unincorporated actin by incubation buffer, the initially spherical actin-containing ves- washing with an excess of the appropriate nonpolymerizing icles assumed markedly asymmetric shapes (Figs. 1 and 2). or polymerizing buffer (10 or more volumes), using Centrex Actin polymerization was confirmed by a decrease of actin concentrators (1-gm pore diameter, Schleicher & Schuell). mobility as measured by FPR (Fig. 1) (6, 20). There was There were no detectable differences in the shapes ofvesicles considerable fluorescence recovery of actin monomer in prepared by dilution or by using Centrex concentrators. vesicles incubated in G buffer (Fig. la). Actin was immobi- FPR. FPR was performed as described (4, 6, 20) with a X 16 lized by formation oflong filaments (20) in vesicles incubated Zeiss Neofluar objective (numerical aperture, 0.35), which in G buffer with 100 mM KCl (Fig. lb). In control experi- gives a 4-gm Gaussian laser spot. Only vesicles with diameter ments, vesicle-encapsulated fluorescein-labeled dextran mol- >10 ,m were used for FPR experiments. ecules diffused freely regardless ofwhether the vesicles were Microscopy. Samples were viewed on a Zeiss universal incubated in G buffer with 100 mM KCI or G buffer without microscope by using a x40 Zeiss Plan objective (numerical the KCI (Fig. 1 c and d). Although the FPR measurements aperture, 0.65). Images were collected with a Dage ISIT showed qualitatively large differences in the diffusion rates of camera (MTI 66), averaged, and digitized with a Grinnell actin under polymerizing and nonpolymerizing conditions, GMR-274 image processor. Microphotographs were taken they did not readily yield quantitative values for the diffusion from digitized images shown on a Panasonic WV-5300 video coefficient and fraction ofmobile actin molecules because the monitor. Scanning confocal microscopy was performed with diameters of the vesicles were not large compared to the a Zeiss Axioplan microscope equipped with a Bio-Rad MRC- diameter ofthe laser spot. Thus, the apparent mobile fraction 500 laser scanning instrument. A x 100 Zeiss Neofluar ob- was reduced because of depletion of the pool offluorophore, jective (numerical aperture, 1.30) was used, and images were and the apparent diffusion rate was increased because of the averaged by using the Kalman model (n = 20) available in the small available volume. Although these effects have been MRC-500 software. Optical section thickness was approxi- quantitatively analyzed for planar systems (21), the analysis mately 0.7 Am. is much more complex for three-dimensional systems and has a b 0 0 (N o- C%4 00 0 o

0 _ 4 g) Cct - 0O C C. 0UZ 0-0 0.

FIG. 1. FPR traces and sample 0 10 20 30 40 o io 20 30 40 phase-contrast photomicrographs of ac- Time (sec) Time (sec) tin- and dextran-containing vesicles. The actin concentration is 48 ,uM throughout. C d (a) Actin-containing vesicles incubated in G buffer, showing recovery of fluores- 0 0. o cence signal in FPR trace after photo- (N ical morphology (Inset). (b) Actin- 0 to *;:--::-:---:--:: containing vesicles incubated in buffer bo : --::::-o ~with--100.::~^mM KC1;----:FPR- ~traces. no longer ._ -O show fluorescence recovery, and theves- c 0 icles are markedly deformed (Inset). (c) C Fluorescein-congugated dextran-con- o . _ taining vesicles (fluorescein-labeled dex- 0 10 20 .. 30 40 tran at 2 mg/ml) incubated in G buffer, to Tie(:c showing recovery of signal and spherical morphology (Inset). (d) Fluorescein- dextran-containing vesicles incubated in 0 0 1 20 G buffer with 100 mM KCI, showing C)1 20 30 40 recovery of signal in FPR and spherical Time (sec) morphology (Inset). Downloaded by guest on September 26, 2021 -.''..''.*ihI Biochemistry: Cortese et al. Proc. Natl. Acad. Sci. USA 86 (1989) 5775 b

c d

e f

FIG. 2. Phase-contrast (a, c, e, g, and i) and fluorescence digital-intensified (b, d, h, and j) images of actin- and actin/gelsolin-containingf, vesicles al- 9 h lowed to equilibrate in buffer with 100 mM KCl for 1 hr as described in Fig. 1. The actin concentration is 48 AM throughout. (a and b) Vesicles containing actin incubated in G buffer are spherical. (c and d) Vesicles containing actin incu- bated in G buffer with 100 mM KCI have a deformed contour and also a heteroge- neous distribution of actin. (e and ff) Vesicles containing actin and 24 nM gelsolin (gelsolin/actin molar ratio of J 1:2000) have a deformed contour and heterogeneous distribution of actin. (g and h) Vesicles containing actin and 0.24 ALM gelsolin (actin/gelsolin molar ratio of 1:200) have a deformed contour and a homogeneous distribution of actin. (i and j) Vesicles containing actin and 0.96 ,M gelsolin (gelsolin/actin molar ratio of 1:50) are spherical, and the distribution of actin is homogeneous. (Bars = 2 ,um.) not been attempted for this work. A further problem for Only vesicles containing actin and able to equilibrate K+ quantitative analysis is a lensing effect due to the high vesicle across the membrane responded to potassium-induced poly- curvature. Nevertheless, the FPR measurements that dem- merization conditions by assuming nonspherical shapes. onstrated immobilization of actin at high K+ concentrations Hence, an osmotically generated force was not responsible showed that actin is extensively polymerized under these for the deformation. Rather, formation of actin filaments conditions. deformed the vesicles. Actin Polymerization Deforms Vesicles. Polymerization of Actin-Binding Proteins Affect Liposome Deformation. The encapsulated actin extensively deformed vesicles (Fig. lb magnitude of vesicle deformation depended on the length of Inset and Fig. 3 a and b). Vesicles observed at 5 min were the actin filaments. Filament length was controlled by mixing undeformed. Deformation developed slowly over the next 30 gelsolin, an actin-binding protein that nucleates filament min (data not shown). Actin-containing vesicles incubated in formation in defined ratios with the actin (22). Under our G buffer (Fig. la Inset) and vesicles containing equivalent conditions (especially where the gelsolin/actin ratio is large), concentrations of fluorescein-labeled dextran or bovine se- self-nucleation by the actin is slow compared to nucleation by rum albumin incubated in G buffer with or without 100 mM gelsolin-actin complexes. Hence, each gelsolin molecule KCI (Insets of Fig. 1 c and d) did not deform. Replacement should ideally nucleate a filament, and the mean length of of K+ by a comparable quantity of Na', which neither these filaments should equal the actin/gelsolin ratio divided supports actin polymerization nor is transported by valino- by 370 monomers per ttm (23). Therefore, the mean length of mycin, left the vesicles undeformed (not shown). Actin- our gelsolin/actin filaments was varied between 0.07 Ium containing vesicles made without valinomycin had larger (1:25) and 5.4 tum (1:2000). It is important to note that these sizes and were insensitive to the presence of 100 mM KCl. are mean lengths, and the actual distribution of filament Downloaded by guest on September 26, 2021 5776 Biochemistry: Cortese et al. Proc. Natl. Acad. Sci. USA 86 (1989) ib

c

FIG. 3. Confocal images of vesicles containing actin (48 AM) and actin-binding proteins, equilibrated with G buffer with 100 mM KCL. (a and b) Vesicles containing actin, deformed and with an inhomogeneous distribution offluorescence. (Bar = 2 jum.) (c) Vesicle containing actin and 0.96 ,M gelsolin (gelsolin/actin molar ratio is 1:50), spherical and with a homogeneous fluorescence distribution. (Bar = 1 ,um.) (d) Vesicle containing actin and 9.6 AM filamin (approximated filamin/actin molar ratio is at least 1:10, considering encapsulation efficiency problems) (9), ellipsoidal and with a peripheral distribution of fluorescence. (Bar = 1 jum.) Images obtained by phase-contrast and conventional fluorescence microscopy on similar samples (as described in the legend for Fig. 2) experienced the same changes in shape (not shown). lengths is quite broad and right-skewed (22). Vesicles con- 3d). In these experiments the initial molar ratio of filamin to taining short actin filaments, obtained by mixing gelsolin and actin was 1:5, but the ratio in the vesicles may have been actin at molar ratios between 1:25 and 1:50, were spherical somewhat lower because of the molecular weight depen- after polymerization (Fig. 2 i and j and 3c). Increasing the dence of encapsulation efficiency (14). lengths ofthe filaments by usinggelsolin/actin ratios between 1:200 and 1:1000 produced nonspherical vesicles (Fig. 2 e-h). At a gelsolin/actin ratio of 1:2000, there was substantial DISCUSSION deformation (Fig. 2 c and h). The transition between spherical Evidence from a variety of sources indicates that cellular and irregular shapes occurred sharply between gelsolin/actin motility and resistance to deformation are dominated by the ratios of1:100 and 1:200, which corresponds to mean filament underlying cytoskeleton (1, 3, 24). Thus, local changes in cell lengths of 0.27 and 0.54 Am, respectively. In addition to its shape and viscoelasticity result from changes in the organi- effect on vesicle shape, the length of the filaments also zation and extent ofpolymerization ofcytoskeletal filaments, influenced their distribution within the vesicles. As the especially actin . In the reconstituted model lengths of the filaments increased, their distribution became system that we have developed, the compositions ofboth the more nonuniform. In contrast to the sharp transition of enveloping membrane and the encapsulated cytoskeleton are vesicle shape, however, the density heterogeneity dimin- far simpler than their cellular counterparts. Nevertheless, as ished gradually as filaments shortened until the distribution of we have shown, the model system does provide useful actin appeared to be homogeneous for gelsolin/actin > 1:50 information about the effects ofchanges offilament length on (mean length c 0.14 ,Am; Fig. 2f, h, andj). Observations by filament organization and on vesicle shape. It also provides scanning confocal microscopy, which provided a narrower a simple approach for demonstrating the role of specific depth of focus, confirmed these findings (Fig. 3). The areas actin-binding proteins in the control ofcell shape and filament of increased filament density were found throughout the organization. Of course the approach is amenable to further vesicles suggesting that gradients of K+ were probably not elaboration, and we expect that it also will be useful for responsible. Moreover, K+ gradients within the vesicles studies of other aspects of cytoskeletal function. should be rapidly dissipated (within about 25 ms) because of When sodium is substituted for potassium at equal osmo- diffusion. larity and when ionophore is omitted, the vesicles do not A filament-crosslinking protein, filamin, had a different deform. Hence the osmotic pressure difference between the effect on vesicle shape. Vesicles containing actin and filamin buffers with and without 100 mM KCI cannot explain the showed a smoother surface than those lacking filamin (Fig. observed vesicle deformation. The role of filament polymer- Downloaded by guest on September 26, 2021 Biochemistry: Cortese et al. Proc. Natl. Acad. Sci. USA 86 (1989) 5777 ization and organization in driving vesicle deformation is We thank Dr. Dorothy Schafer, who assisted us with the confocal further confirmed by the effects of gelsolin, which alters the microscopy experiments; Drs. Michael Sheetz, Robin Michaels, and length distribution of the actin filaments, and filamin, which Tony Pryse for valuable discussions; and John Cooper for a critical crosslinks the filaments. reading of the manuscript. This work was supported by National Long actin filaments not only spontaneously deformed Institutes of Health Grant GM38838 (to E.L.E.), National Research Service Award-Medical Scientist Grant GMO 7200 (to B.S.), Na- vesicles but also assumed a dramatically heterogeneous tional Institutes of Health Grant DK13332 (to C.F.), National Sci- distribution within the vesicles. Heterogeneity in actin gels ence Foundation Grant DMB-8610636 (to E.L.E.), and support for has already been observed on much larger scales (4, 25), but the confocal microscope by the Lucille P. Markey Charitable Trust. our results demonstrate spontaneous domain formation at lower concentrations of actin and at sizes comparable to 1. Elson, E. L. (1988) Annu. Rev. Biophys. Biophys. Chem. 17, animal cells. A possible explanation of the observed anisot- 397-430. ropy would be discrete nucleation sites for polymerization. 2. Alberts, B., Bray, D., Lewis, J., Raff, M., Roberts, K. & Watson, J. (1983) Molecular Biology ofthe Cell (Garland, New Heterogeneous nucleation of actin polymerization at mem- York), pp. 550-611. brane defects or microphase separations could nucleate small 3. Stossel, T. P., Janmey, P. A. & Zaner, K. S. (1987) in Cyto- groups ofparallel filaments which could act as primers for the mechanics, eds. Bereiter-Hahn, J., Anderson, 0. R. & Reif, formation ofa few larger filament bundles (4) or domains (25), W. E. (Springer, Berlin), pp. 131-154. perhaps preferentially oriented along the membrane (26), 4. Cortese, J. D. & Frieden, C. (1988)J. Cell Biol. 107, 1477-1487. which could then deform a vesicle (27). Actin filaments 5. Pollard, T. D. & Cooper, J. A. (1986) Annu. Rev. Biochem. 55, formed in the presence of gelsolin would probably form 987-1035. numerous shorter and more randomly oriented bundles that 6. Axelrod, D., Koppel, D. E., Schlesinger, J., Elson, E. L. & would be insufficiently anisotropic to deform a vesicle. That Webb, W. (1976) Biophys. J. 16, 1055-1069. the nonuniformity of actin density depends weakly on fila- 7. Small, J. V. (1982) in Embryonic Development. Part B: Cellular ment length whereas vesicle shape depends strongly indi- Aspects, eds. Burger, M. M. & Weber, R. (Liss, New York), cates, however, that the relation between these effects is not pp. 341-358. 8. Spudich, J. A. & Watts, S. (1971) J. Biol. Chem. 246, 4855- simple. The observed heterogeneity of actin density could 4871. result from a thermodynamic collapse ofthe actin gel (28, 29). 9. Maclean-Fletcher, S. & Pollard, T. D. (1980) Biochem. Bio- This occurs when the balance offorces driving gel expansion phys. Res. Commun. 96, 18-27. (i.e., gel osmotic pressure) and gel contraction (i.e., inter- 10. Tait, J. F. & Frieden, C. (1982) Arch. Biochem. Biophys. 216, polymeric attraction) shifts towards contraction. Then, the 133-141. volume of the gel can be greatly reduced (and its density can 11. Cooper, J. A., Bryan, J., Schwab, B., III, Frieden, C., Loftus, be correspondingly increased in some areas, producing over- D. J. & Elson, E. L. (1987) J. Cell Biol. 104, 491-501. all a nonuniform distribution of actin filaments) by a small 12. Feramisco, J. R. & Burridge, K. (1980) J. Biol. Chem. 255, change in an external variable such as pH or ion concentra- 1194-1199. tion. Gel-collapse theory is consistent with the filamin re- 13. Deamer, D. & Bangham, A. D. (1976) Biochim. Biophys. Acta sults. Filamin, which crosslinks actin, might have been 443, 629-634. expected to increase vesicle deformation by increasing the 14. Szoka, F., Jr., & Papahadjopoulos, D. (1980) Annu. Rev. of the actin gel that deforms the Biophys. Bioeng. 9, 467-508. mechanical strength (30) 15. Jain, M. K. (1983) in Membrane Fluidity in Biology, ed. Aloia, vesicle. However, gel-collapse theory predicts that increases R. C. (Academic, New York), pp. 1-35. in the elastic modulus of the gel would resist collapse. Our 16. Papahadjopoulos, D. & Watkins, J. C. (1967) Biochim. Bio- results showed that inclusion offilamin gave a more isotropic phys. Acta 135, 639-665. actin distribution and less vesicle deformation than actin 17. Deamer, D. W. (1978) Ann. N. Y. Acad. Sci. 308, 250-258. alone (Fig. 3d). Gel collapse would enable a small change in 18. Laliberte, A. & Gicqauaud, C. (1988) J. Cell Biol. 106, 1221- the conditions (ion concentration, polymer concentration and 1227. number, and polymer elasticity) to drive important filament 19. Tanaka, T., Fillmore, D., Sun, S.-T., Nishio, I., Swislow, G. & rearrangements (31). Shab, A. (1980) Phys. Rev. Lett. 45, 1636-1639. The abrupt effect of small changes in filament length on 20. Tait, J. F. & Frieden, C. (1982) Biochemistry 21, 3666-3674. 21. Angelides, K. J., Elmer, L. W., Loftus, D. & Elson, E. (1988) vesicle shape and heterogeneity of actin density suggests that J. Cell Biol. 106, 1911-1925. corresponding changes in a cell could affect its shape and 22. Janmey, P. A., Peetermans, J., Zaner, K. S., Stossel, T. P. & cytoskeletal organization. For example, if actin filament Tanaka, T. (1986) J. Biol. Chem. 261, 8357-8362. lengths in a cell were in an appropriate range, then the 23. Hanson, J. & Lowy, J. (1963) J. Mol. Biol. 6, 46-60. activation or inactivation of a small number of capping pro- 24. Bray, D. & White, J. G. (1988) Science 239, 883-888. teins (e.g., gelsolin) could change the filament lengths suffi- 25. Buxbaum, R. E., Dennerll, T., Weiss, S. & Heidemann, S. R. ciently to trigger a localized change in cell shape such as the (1987) Science 235, 1511-1514. extension of a protopod. Spontaneous reorganization of fila- 26. Lekkerkerker, H. N. W., Coulon, P., Van Der Haegen, R. & ments could be stabilized by actin-crosslinking proteins such Deblieck, R. (1984) J. Chem. Phys. 80, 3427-3433. 27. Mozzarelli, A., Hofrichter, J. & Eaton, W. A. (1987) Science as filamin (30) or a (32). It is likely, however, that 237, 500-506. additional processes such as changes in gel osmotic pressure 28. Tanaka, T. (1981) Sci. Am. 244 (1), 124-138. and interactions with the membrane are important (19, 31, 33, 29. Mark, J. E. & Erman, B. (1988) in Rubberlike Elasticity: A 34). These could also depend critically on filament length (35). Molecular Primer (Wiley, New York), pp. 127-132. Our results demonstrate that actin polymerizationper se can 30. Brotschi, E. A., Hartwig, J. H. & Stossel, T. P. (1978) J. Biol. drive biological membranes to change shape. We have shown Chem. 253, 8988-8993. that interactions among actin filaments that vary with filament 31. Stokke, B. T., Mikkelsen, A. & Elgsaeter, A. (1986) Eur. length are sufficient both to cause the vesicles to change shape Biophys. J. 13, 219-233. and to cause an inhomogeneous distribution of actin in the 32. Sato, M., Schwarz, W. H. & Pollard, T. D. (1987) Nature (London) 325, 828-830. vesicles. Similar thermodynamic forces could influence dy- 33. Stokke, B. T., Mikkelsen, A. & Elgsaeter, A. (1986) Eur. namic cytoskeletal processes in living cells. These model Biophys. J. 13, 203-218. systems should also be quite useful for probing the effects of 34. Elgsaeter, A., Stokke, B. T., Mikkelsen, A. & Branton, D. ion fluxes and actin-binding proteins on the assembly and (1986) Science 234, 1217-1223. organization of filament systems and the mechanical conse- 35. Ito, T., Zaner, K. S. & Stossel, T. P. (1987) Biophys. J. 51, quences of filament interactions on a cellular scale. 745-753. Downloaded by guest on September 26, 2021