Aggregation and Calcium-induced Fusion of Phosphatidylcholine Vesicle-Tubulin Complexes*

Insertion of tubulin into the bilayer of dipalmitoyl phosphatidylcholine vesicles at the phase transition results in the formation of stable vesicle-tubulin com- plexes (Klausner, R. D., Kumar, N., Weinstein, J. N., Blumenthal, R., and Flavin, M. (1981) J. Biol. Chem 256, 5879-5885). These complexes aggregated when maintained below phase transition for 10-20 min. Addition of millimolar concentrations of Ca2+, Mn2+, Zn2+, and Co2+, but not M&+, caused the vesicle-tubulin com- plexes to fuse into larger structures as shown by (a) electron microscopy, (b) increased trapped volume, and (c) changes in resonance energy transfer between two fluorescent lipid probes incorporated into the same vesicle. There was no loss of internal aqueous contents from the vesicle-tubulin complexes during Ca2'-in- duced fusion. Anti-tubulin drugs had no effect on the aggregation or fusion, and vesicle-bound tubulin did not associate with microtubules when tubulin was assembled in vitro. Trypsin-treated vesicle-tubulin com- plexes were incapable of supporting Ca2'-induced fusion. This system provides a model for transition temperature, i.e. 40 “C for 2 min. We have shown earlier that tubulin is inserted into the bilayer of these vesicles at the phase transition to form stable complexes (6). The trapped volume of vesicles (volume of the total internal aqueous compartment) was determined by preparing and incubating the vesicles at 40 “C in the presence of 20 mM carboxyfluorescein. The vesicles and vesicle-tubulin complexes were passed over a Seph- adex G-25 (PD-10, Pharmacia) column, and the total internal volume of the vesicle per mol of phospholipid was determined by spectrofluor- imetric measurement of carboxyfluorescein. The Ca*+-induced “fused” vesicles were too large to pass through the G-25 column. They were centrifuged at 100,ooO X g in an Airfuge (Beckman Instruments). The pellet was washed seven times with buffer contain- ing 150 mM NaCl, 10 mM Hepes, and 1 mM EGTA, pH 7, dissolved with Triton X-100, and carboxyfluorescein and lipid were measured. The volume of the aqueous compartment of vesicles relative to total volume was determined by measuring the ratio between fluorescence before and after separating the vesicles from the medium containing 20 mM carboxyfluorescein. The phospholipid concentration (mol/ liter) was measured after the separation. The trapped volume (liters/ mol) was then determined by dividing the fluorescence ratio by the phospholipid concentration.

sized on rough microsomes, incorporated into microsomal membranes, and translocated to the plasma membrane as a component of vesicles.
We have recently shown (6, 7 ) that purified tubulin can be incorporated into dipalmitoyl phosphatidylcholine vesicles at the lipid phase transition temperature, without any requirement for detergent or sonication, and results in the formation of stable complexes. Similar results had been reported by Caron and Berlin (8) using dimyristoyl phosphatidylcholine vesicles. The insertion process is accompanied by structural perturbations of both the lipid bilayer and the tubulin (6, 7 ) .
In this paper we examine the properties of the vesicle-tubulin complexes. These complexes aggregated even in the absence of divalent cations; in the presence of millimolar CaZ+ or Mn2+ but not Mg2+, they fused to form much larger closed structures without release of the water-soluble internal marker, carboxyfluorescein. We have also compared free and vesicle-bound tubulin for their ability to bind colchicine and MAPs' or to be tyrosinolated and detyrosinolated by specific enzymes.

MATERIALS AND METHODS
Preparation of Tubulin-Microtubule protein was purified from freshly obtained bovine brains by three cycles of assembly and disassembly without glycerol, according to the procedure of Asnes and Wilson (9). Tubulin2 was further purified by phosphocellulose chromatography as described elsewhere (6) and stored in buffer containing 100 mM K'-2-(N-morpho1ino)ethanesulfonic acid, 1 m M MgSO.,, 1 m M EGTA, 2 m M dithiothreitol, and 0.1 m M GTP, pH 6.8, at -70 "C. MAPS which remain bound in the column were subsequently eluted with 1.0 M KC1 in the above buffer, concentrated using an Amicon PM-30 filter, and stored at -70 "C. Protein was determined according to the method of Lowry et al. (10) using bovine serum albumin as standard. This refers to tubulin purified by phosphocellulose chromatography. 3 X tubulin refers to three assembly cycles purified tubulin.

15137
transition temperature, i.e. 40 "C for 2 min. We have shown earlier that tubulin is inserted into the bilayer of these vesicles at the phase transition to form stable complexes (6).
The trapped volume of vesicles (volume of the total internal aqueous compartment) was determined by preparing and incubating the vesicles at 40 "C in the presence of 20 mM carboxyfluorescein. The vesicles and vesicle-tubulin complexes were passed over a Sephadex G-25 (PD-10, Pharmacia) column, and the total internal volume of the vesicle per mol of phospholipid was determined by spectrofluorimetric measurement of carboxyfluorescein. The Ca*+-induced "fused" vesicles were too large to pass through the G-25 column. They were centrifuged at 100,ooO X g in an Airfuge (Beckman Instruments). The pellet was washed seven times with buffer containing 150 mM NaCl, 10 mM Hepes, and 1 mM EGTA, pH 7, dissolved with Triton X-100, and carboxyfluorescein and lipid were measured. The volume of the aqueous compartment of vesicles relative to total volume was determined by measuring the ratio between fluorescence before and after separating the vesicles from the medium containing 20 mM carboxyfluorescein. The phospholipid concentration (mol/ liter) was measured after the separation. The trapped volume (liters/ mol) was then determined by dividing the fluorescence ratio by the phospholipid concentration.
Centrifugation Assay for Aggregation/Fusion of Vesicles or Vesicle-Tubulin Complexes-This assay made use of the change in sedimentation constant as a result of aggregation and/or fusion of vesicles. Routinely, vesicles alone or vesicle-tubulin complexes were incubated for 15-30 min below the phase transition temperature of DPPC (usually at 28-30 "C) in tubes (7 X 50 mm; cellulose acetate butyrate). The assay buffer was 100 mM K+-2-(N-morpholino)ethanesulfonic acid (pH 6.8) containing 0.5 mM MgS04 and 0.75 mM EGTA. At the end of incubation, tubes were centrifuged (using adaptors no. 408) at 40,000 X g for 20 min at 25 "C (Sorvall Rotor SS-34). Supernatants were removed for counting. The bottoms of the tubes were cut off and dropped into scintillation vials for counting. The recovery of total radioactivity (pellet and supernatant) was >98%.
Results are expressed as the ratio of radioactivity in the pellet/ total radioactivity. The actual amount pelleted varied between 20 and 65% (at a tubulin to vesicle molar ratio of 101) with several vesicle and/or tubulin preparations. In general, this variability was seen only with the pelleting of vesicle-tubulin complexes in the absence of added Ca2+. However, any single preparation gave extremely reproducible results.
Resonance Energy Transfer between Lipid Probes-Mixing of lipids was assayed according to the procedure developed by Pagano and his co-workers (11) involving resonance energy transfer between two fluorophores incorporated into the vesicle bilayer. The energy donor NBD and the energy acceptor rhodamine are coupled to the free amino group of phosphatidylethanolamine to form N-NBD-PE and N-Rh-PE (11). The two lipid probes were obtained from Avanti Polar Lipids. As shown by Struck et al. (11) when both fluorescent lipids are in lipid vesicles at appropriate surface densities (ratio of fluorescent lipid to total lipid), efficient energy transfer is observed.
Steady state emission and excitation spectra were obtained by using a Perkin-Elmer MPF 44b spectrofluorometer, the excitation band slit was at 4 nm, and the emission slit at 10 nm. Following each set of measurements, vesicles were disrupted with Triton X-100 (0.1% final concentration). This treatment eliminated energy transfer (see below) and allowed the determination of the concentrations of N-NBD-PE and N-Rh-PE from their emission intensities by using direct excitation. All spectra were obtained at room temperature (22 "C).
Spectra of tubulin-vesicle preparations containing both N-NBD-PE and N-Rh-PE (DPPC:N-NBD-PE:N-Rh-PE = 981:1), excited at 450 nm, show emission maxima at 530 nm and 585 nm (see Fig. 5 under "Results"). Essentially all of the fluorescence at 530 nm comes from N-NBD-PE, whereas the fluorescence at 585 nm arises from fluorescence energy transfer between the donor and acceptor pair. The efficiency of energy transfer (quenching of the energy donor) in such samples is defined by the relationship (12) where F is the fluorescence at 530 nm in the presence of N-Rh-PE and Fo is the fluorescence at 530 nm in the absence of N-Rh-PE. Struck et al. (11) for phosphatidylserine vesicles containing the same Electron Microscopy-For electron microscopy, DPPC vesicles and vesicle-tubulin complexes with or without Ca2+ were negatively stained within an hour of the time that the vesicle-tubulin complexes were prepared. Uranyl acetate and phosphotungstic acid in several protocols were tested for studying the vesicles and vesicle-tubulin complexes in the electron microscope. Procedures included mixing the vesicles with stains before application to the grids or application of the vesicles to the grids followed by blotting and/or various washes before application of the stain. (The problems and advantages of these techniques are considered in more detail under "Discussion".) After several stains and protocols had been tested, the following procedure was adopted. All manipulations were done at room temperature which is below Tc for DPPC. A drop (about 8 p l ) of each sample was pipetted onto carbon and Formvar-coated grids. After 1 min or less, the grids were blotted but not allowed to dry. A drop of 2% phosphotungstic acid, pH 6.5, was applied, and after 1 min or less the grids were blotted, allowed to dry, and examined using a Phillips 400 microscope at 60 kV.

Aggregation of Vesicle-Tubulin Complexes
Fig. 1 shows the typical turbidity changes (indicated by changes in 90" light scattering at 470 nm) of DPPC vesicles upon addition of various concentrations of tubulin at 40 "C. Without tubulin, there was no change in the light scattered by vesicles, whereas increasing amounts of tubulin resulted in an increase in the scattering, probably as a result of aggregation of the small unilamellar DPPC vesicles. There was very close correspondence in the time course for the turbidity increase and the release of encapsulated carboxyfluorescein reported previously (6). The change in turbidity takes place only when tubulin is incubated with vesicles around 37-40 "C (which is also the transition temperature), A likely interpretation of the observed increase in the turbidity is that tubulin promoted the aggregation of vesicles under conditions in which it is inserted into the lipid bilayer (6). A variety of other soluble proteins such as hemoglobin, immunoglobulin, albumin, and ovalbumin do not produce an effect similar to that of tubulin on these vesicles. Formation of larger structures was also indicated by the relative increase in the sedimentation of vesicles as shown in Table I  lipid) containing carboxyfluorescein were incubated with various amounts of t u b u l i in the cuvettes. The numbers in the figure indicate molar ratio of tubulin to vesicles (assuming 4500 lipid per vesicle). Temperature of the cuvette holder was maintained by recirculating water at 42 "C. Ninety degree light scattered at 470 run was determined using an Aminco-Bowman spectrofluorometer with 2-mm slits and a two-channel Y-t recorder. Tubulin was added at time zero (below phase transition temperature). Experiments were performed in the presence of 1 m~ EGTA. Tubulin alone did not show any light scattering changes.

Effect of varying tubulin/vesicle ratio on the pelleting of vesicles
Vesicle-tubulin complexes of varying tubulin to vesicle ratios were prepared in buffer containing 0.75 m M EGTA. They were incubated at 30 "C for 15 min before centrifugation (40,000 X g for 30 min) to determine pelleting of vesicles, as described under "Materials and Methods." vesicle-tubulin complexes + 2.5 mM CaClz. All the incubations and centrifugation were at 30 "C. Numbers in parentheses represent the results obtained when incubation and centrifugation were done at 0 "C. alone nor was there any significant effect on the pelleting when tubulin and vesicles were mixed below Tc. When tubulin was inserted into the lipid bilayer at the endothermic Tc approximately 40 to 65% of the vesicles could be pelleted in the absence of Ca2+. Addition of Ca2+ to the vesicle-tubulin complexes, however, caused nearly complete pelleting of the vesicles (Fig. 2). The effect of Ca2+ was dependent upon the final concentrations added to vesicle-tubulin complexes. As shown in Fig. 3, maximum pelleting occurred above 2.5 mM concentration. Since these assays were performed in buffer containing 0.75 mM EGTA, the threshold concentration of free Caz+ for maximum pelleting was above 1.0 mM. The inset of Fig. 3 also shows that Mg2' was without any effect, whereas Mn2+, Zn2+, and Co2+ enhanced the pelleting of the vesicletubulin complexes. Effectiveness of various cations tested was Evidence for Ca2+-induced Fusion Electron Microscopy- Fig. 4 is a series of electron micrographs prepared by the technique described under "Materials and Methods." It illustrates the effect of Ca2+ on vesicletubulin complexes. The population of DPPC vesicles eluted from the Sepharose 4B column contained primarily single spherical vesicles 18-30 nm in diameter. Among these vesicles, there was a small number of larger vesicles (up to 10 nm in diameter) each of which appeared to contain several of the small vesicles (Fig. 4A). DPPC vesicles incubated with 5 mM Ca2+ were indistinguishable from those without Ca2' (Fig.   4B). DPPC vesicles into which tubulin had been incorporated by passage through Tc showed marked clumping, but there was also a substantial population of free small vesicles. Within the clumps, individual small vesicles of the original size were clearly distinguishable (Fig. 4C). Fig. 4C was selected from an area including some of the largest aggregates found. Tubulincontaining vesicles incubated in the presence of 5 mM Ca2+ completely gathered into clumps with no free vesicles apparent. Within these clumps in addition to some distinguishable small vesicles there were also vesicles of a variety of larger sizes as well as much larger structures apparently composed of sheets of membrane folded into topologically complex forms (Fig. 40). Fig. 4 0 was selected from an area showing the highest proportion of distinguishable small vesicles in addition to the larger membrane forms. The most obvious interpretation is that the larger vesicles and complex membrane structures are formed by fusion of the small vesicles. The differences between vesicles and vesicle-tubulin complexes with or without ea2+ were obvious and reproducible.
Trapped Volume-Small unilamellar vesicles as well as vesicle-tubulin complexes at a 30:l ratio (tubulin to vesicle) were formed in 150 mM NaC1-10 m~ Hepes buffer, pH 7.0, containing 20 m~ carboxyfluorescein and 1 mM EGTA. Fusion was induced by adding Ca2+ to a final concentration of 10 mM to the vesicle-tubulin complexes in a medium containing 20 mM carboxyfluorescein. The trapped volume was measured after separation of the vesicles from the medium. As shown in Table 11, vesicle-tubulin complexes had a 2-fold increase in trapped volume over the original vesicles, and Ca2+-induced fusion produced a 10-fold increase in trapped volume. This result indicates that at least some of the large structures seen in Fig. 4

Trapped volume of vesicles and vesicle-tubulin complexes with and
without Ca'+ The vesicles and vesicle-tubulin complexes were prepared in 20 mM carboxyfluorescein, 150 mM NaCI, 10 mM Hepes, 1 mM EGTA, pH 7. Trapped volume was measured after separation of vesicles or vesicletubulin complexes from the incubation medium. Tubulin to vesicle ratio was 301. In Fig. 6B we have plotted the energy transfer efficiency calculated according to Equation 1 as a function of Ca2+ concentration. At about 1 mM Ca2+ there is a 2-fold decrease in transfer efficiency, indicating that on the average each fluorescent vesicle has fused with an unlabeled vesicle. The Ca2+ dependence shown in Fig. 6 is quite similar to the Ca2+ dependence of pelleting shown in Fig. 1. In accordance with the divalent cation specificity of pelleting we also see that Mg2+ had little effect on energy transfer, whereas the effect of Mn2+ was similar to that of Ca".
When the probe-containing vesicles were incubated with Ca2+ in the absence of pure vesicles there was no change in energy transfer, indicating that the effect was not due to, for instance, a change in the radius of curvature of the vesicles. However, when we incubated probe-containing vesicle-tubulin complexes with DPPC vesicles without tubulin we did see Ca2+-dependent changes in energy transfer (Fig. 6). This be mixed immediately after formation with the unlabeled vesicles to produce efficient Ca'+-dependent energy transfer changes. When they were allowed to stand for a while in the absence of Ca2+ they presumably aggregated. When those were mixed with unlabeled vesicle-tubulin complexes there was no marked change in energy transfer efficiency upon addition of Ca" (not shown). The aggregated labeled complexes had presumably fused with each other and not with  Other controls (not shown) which did not produce Ca*+dependent energy transfer changes include mixing vesicles with Ca" without tubulin and mixing vesicles with tubulin (without phase transition) and Ca". Ca "-induced Fusion of Vesicle-Tubulin Complexes was Nonleaky-An important question about vesicle-vesicle fusion is whether the fusion process is accompanied by leakage of vesicle contents. This is the case with CaZf-induced fusion of phosphatidylserine vesicles (19). In order to assess leakage we formed vesicle-tubulin complexes by bringing DPPC vesicles containing 120 mM carboxyfluorescein to Tc in the presence of tubulin at a tubulin-vesicle ratio of 30~1. We then immediately rechromatographed the complexes on a Sephadex G-25 (PD-10) column to remove dye released during complex formation. All this was done in a buffer containing 1 mM EDTA. The ratio between the fluorescence of the rechromatographed complexes before and after disrupting them with Triton X-100 was 0.23. This is consistent with 30 mM carboxyfluorescein remaining in the vesicles (see Fig. 3 in reference 6). This indicates that about 70% carboxyfluorescein was released from the vesicles during complex formation. Fig. 6A shows that addition of Ca2+ at concentrations which induced efficient fusion (see Fig. 6B) did not cause leakage of carboxyfluorescein from those complexes. The average leakage, probably due to dilution of the complexes into the Ca'+ buffer, was 4.6 + 3.3% (dashed line in Fig. 6A). The leakage at the different Ca*' concentrations was not significantly different from that background level. Further Characterization of Aggregation and Ca'+-induced Fusion-In the next set of experiments we tested whether the pelleting of lipid is also accompanied by a proportional amount of tubulin in the pellets. Results of one such experiment in which we used 'H-DPPC vesicles and tyrosinolated ["'Cltubulin are shown in Table III. There is no difference between tyrosinolated and detyrosinolated tubulins in their interaction with DPPC vesicles (see below). Tyrosinolated ['4C]tubulin was prepared according to the method described elsewhere (13). This modification of tubulin involves addition of a tyrosine residue at the COOH terminus of the (Y subunit, catalyzed by tubulin tyrosine ligase in the presence of ATP. No significant differences in the in vitro microtubule assembly properties so far have been observed between maximally tyrosinolated and detyrosinolated tubulins (14). Tubulin alone under these conditions did not pellet, with or  Table IV show that actin, another cytoskeletal protein which interacts with DPPC vesicles, also mediated the Ca2+-induced sedimentation of vesicles. On the other hand, serum apolipoprotein Al, which also inserts into these vesicles, did not behave like tubulin and actin.
Agents which affect polymerization of tubulin (polylysine, colchicine, and podophyllotoxin) did not have any significant effect on the tubulin-mediated or Ca2'-enhanced aggregation and fusion of vesicles. MAPs reproducibly caused 10-15% inhibition of pelleting of vesicles (Table V). Bovine serum albumin and NaCl had no effect on the vesicle sedimentation with or without added Ca2+. Brief treatment with trypsin abolished the Ca2+-induced sedimentation of tubulin-containing vesicles (Table VI). This result suggests a role for the exposed portion of the tubulin molecule (i.e. that is not buried in the bilayer) in mediating the Ca2+ effect.
To test whether the Ca2+ effect reflected enhanced proteinprotein interaction or whether inserted tubulin could interact with pure lipid, we made vesicle-tubulin complexes with unlabeled lipid and examined the pelleting of 14C-labeled (protein-free DPPC) vesicles. Data in Table VI1 show that Ca2+induced pelleting of vesicles can take place between tubulincontaining and protein-free vesicles. However, it is necessary that tubulin be inserted in at least some participating vesicles.
Finally we have examined whether vesicle-tubulin complexes interact with microtubules during polymerization of tubulin. Results shown in Table VI11 suggest that neither . esicles nor tubulin-containing vesicles bind tightly to microtubb'os. Assembled microtubules were separated from the bulk of unassembled contents by centrifugation through a 50% sucrose cushion. In these experiments more than 60% of the tubulin was recovered in the microtubule pellet, and 10 PM colchicine resulted in 90% inhibition of this polymerization. Radioactive lipid in the pellets using either vesicles or vesicletubulin complexes was the same with or without colchicine and represented only about 6% of the total lipid.
Further Characterization of Tubulin-Vesicle Interaction-Association of tubulin with DPPC vesicles was studied as described earlier (6) by measuring the release of encapsulated dye (carboxyfluorescein) into the medium. The amounts of protein required to cause 50% (of total) release of the dye are shown in Table IX. Three assembly-cycle purified tubulin (3 X tubulin) which has approximately 80% tubulin and 20% MAPs was 9 to 10 times less effective than purified tubulin, and MAPs (up to 3 mg/ml) were totally ineffective in inducing  Further characterization of tubulin-mediated and Ca'+-induced aggregation/fusion of DPPC vesicles Tubulin to vesicle ratio was 101. Vesicle-tubulin complexes were incubated with various substances tested for 20 min before centrifugation to determine pelleting of vesicles. CaZi was added to a concentration of 2.5 mM. (1 mg/ml)

TABLE VI
Trypsin treatment of vesicle-tubulin complexes abolishes Ca2+enhanced aggregation/fusion of vesicles Tubu1in:vesicle ratio was 1O:l. Trypsin concentration was 3 pg/ml. After 10 min at 30 "C, trypsin was inhibited by the addition of soybean trypsin inhibitor (9 pg/ml). Trypsin-treated or freshly prepared vesicle-tubulin complexes were further incubated in the presence of 5 mM CaCI2 for 30 min at 30 "C and processed for the determination of pelleting of vesicles. Soybean trypsin inhibitor alone had no effect.

TABLE VI1
Aggregatwn/fusion can also take place between vesicle-tubulin and protein-free vesicles Tubulin to vesicle ratio was 101 and CaC12 concentration, 2.5 mM. In the sets in which radioactive and nonradioactive vesicles were combined, equal lipid concentrations for both were used. Incubations were done at 30 "C.

Condition
Radioactivity in pellet -CaC12 +CaCla 76 of total Vesicles" 16 11 Vesicles" + vesicles dye -*elease. In fact, addition of MAPs to tubulin during phase transition release resulted in the partial inhibition of release (not shown). There was no difference in the ability of maximally tyrosinolated and detyrosinolated tubulins to cause release of the dye from DPPC vesicles. Tyrosinolated and detyrosinolated tubulins were prepared by incubating tubulin with tubulin tyrosine ligase and carboxypeptidase A as described (13). There was no significant effect of GTP and various anti-tubulin drugs such as colchicine and podophyllotoxin in the phase transition release assay. On the other hand, heat inactivation of the tubulin (90 min at 45 "C) dramatically destroyed phase transition release activity. Since there was no difference between tyrosinolated and detyrosinolated tubulins in their interaction with vesicles, we also VI11 Vesicle-tubulin complexes do not associate with microtubules 3 x tubulin (1.5 mg/ml) was assembled for 45 min at 30 "C in the presence or absence of 10 pM colchicine in a total volume Of 300 pl. (c) and (d) contained 25% tubulin as vesicle-tubulin complex (10 tubulins per vesicle). Polymerized microtubules were collected by centrifugation of the assembled solution through a 50% sucrose cushion (Beckman rotor 65 at 50,000 rpm for 2 h, 25-30 "c). Pellets were redissolved in 0.1 N NaOH and analyzed for total protein and radioactivity. In (c) to (f) total "C counts (['4C]DPPC vesicles) added during polymerization were 7300.  Values are mean f S.D. of 3 to 4 experiments. Tubulin-drug complexes were prepared by incubating tubulin with 2-fold molar excess of drugs at 30 "C for 60 min. Free (unbound) drug was removed by gel chromatography on a Sephadex G-50 column.
Tubulin" 11 f 5 3 X tubulin MAPS Tubulin-colchicine complex tested if there would be differences in the ability of these tubulins to act as substrates for the enzymes which cause these modifications. In several experiments, we did not see any differences between the ability of free and vesicle-bound tyrosinolated tubulin to be detyrosinolated by purified carboxypeptidase (13) or in the ability of detyrosinolated tubulin to be tyrosinolated by tubulin-tyrosine ligase (15).
Vesicle-bound tubulin was equally active in binding colchicine, and the kinetics of binding of [3H]colchicine to free and vesicle bound tubulin was identical.
[3H]Colchicine binding to tubulin was tested according to the procedure described elsewhere (16). Aliquots of tubulin or vesicle-tubulin complexes were incubated in 150 pl of a buffer containing 10 mM NaH2P04, 5.0 mM MgS04, 0.24 M sucrose, and 0.1 mM GTP, pH 6.9, and 1.2 pCi/ml of ['H]colchicine (final concentration 12 p~) at 30 "C for 15,30,60,120, and 180 min. At each time point a 100-pl aliquot was transferred onto a DE-23 column (0.5 ml in Bio-Rad econo-column 731-1110), pre-equilibrated with 15 ml of the same buffer but without sucrose and GTP. The columns were washed immediately with 1.5 ml of buffer. Tubulin was retained in the column. The column contents were quantitatively transferred into scintillation vials. Radioactivity was measured using Aquasol (New England Nuclear). Nonspecific binding assessed by incubating as above but in the presence of 50-fold excess of nonradioactive colchicine was less than 2-3% of total binding. A similarly low value was obtained when [3H]colchicine alone (no tubulin) was identic d y incubated and processed. Tubulin in the form of vesicletubulin complexes bound a significant amount of [32p] MAP2 (up to approximately 1 mol of MAPz per 15 mol of tubulin). [:'*PIMAP2 was prepared by autophosphorylation of 3 x tubulin and heat treatment as described (17,18). Aliquots of tubulin or vesicle-tubulin complexes (approximately 10 tubu1in:vesicle) were incubated at 30 "C for 60 min with varying amounts of ["2P]MAPz and 0.5 mM GTP. Tubes were centrifuged in a n Airfuge (20 p.s.i., 30 min a t 25 "C) and supernatants and pellets were analyzed for radioactivity and protein. The amount of tubulin in pelleted vesicle-tubulin complexes was estimated on the basis of results described in Table 111. Under these conditions, microtubules had bound approximately 1 mol of MAP2 per 9 mol of tubulin, a value similar to the one reported elsewhere (14,19).

DISCUSSION
The studies reported in this paper show that the insertion of tubulin into lipid bilayers of uncharged sonicated small unilamellar vesicles results in aggregation of the vesicles, measured by light scattering and increased pelleting of the vesicles. Simply mixing the protein and vesicles was not sufficient to cause the aggregation of vesicles. In the presence of millimolar Ca'+ or Mnz+ but not Mg2+, there was nearly complete pelleting of the vesicles. The effects of Ca'+ on the DPPC vesicle-tubulin complexes could be related to the fact that tubulin is a Ca2+-binding protein (20). Data in Table V suggest that protein-protein interaction during aggregation of vesicles does not involve polymerization of tubulin as judged by the lack of effect of polymerization inhibitors and that vesicle-tubulin complexes do not associate with microtubules. Results in Table VIII, however, do not rule out the possibility of dissociation of weakly bound vesicles from microtubules during centrifugation.
Actin, another component of the cytoskeletal system when incubated with lipid vesicles and passed through the phase transition also promoted increased pelleting of vesicles, which was enhanced by Cap+. We have recently observed that clathrin, a component of coated vesicles can produce fusion of dioleoyl phosphatidylcholine vesicles at pH below 6.:' Apolipoproteins, on the other hand, while they form vesicular recombinants with DPPC by passage through the phase transition, do not promote increased pelleting.
Based on pelleting and light-scattering data, it is not possible to differentiate between aggregation and fusion of vesicles. Electron microscopy, however, shows that inserted tubulin results in the aggregation of vesicles, and only the addition of divalent cations induces membrane fusion. Although these results with electron microscopy were clear and reproducible, a number of authors have pointed out a variety of artifacts that can be produced by negative staining of phospholipid preparations (see for example reference 21); and it is perhaps warranted to discuss the question of the reliability of our results in more detail. The results are internally consistent and show clear differences between the controls: vesicles alone, vesicles with Ca2+, and tubulin-vesicle complexes without ea2+ which did not fuse. If the final fusion of vesicles were caused by phosphotungstic acid it would have been a phosphotungstic acid-induced fusion dependent upon the insertion of tubulin into the DPPC vesicles and requiring calcium. In our protocol the effects of the stain on aggregation and fusion of the vesicles were minimized by applying the vesicles or vesicle-tubulin complexes to grids and blotting or washing off the excess fluid before applying the stain. This should have left primarily vesicles that were adsorbed to the substrate to be outlined by the stain. (Washing the grids bearing the vesicles or complexes with water before application of the stain or rinsing away floating vesicles in a stream of stain gave essentially the same results as those shown in Fig. 4). On the R. Blumenthal, M. Henkart, and C. Steer, unpublished observations.
other hand, when DPPC vesicles were mixed with phosphotungstic acid in suspension, we observed stacking of the vesicles as described by Melchior et al. (21). Uranyl acetate which is generally considered the negative stain of choice for studies of microtubules was not useful for this study because the divalent cation UOZ2+ is known in many systems to mimic the effects of CaZt which was the object of these experiments. In fact, with uranyl acetate staining there was extensive reorganization of the tubulin-vesicle complexes with or without calcium while the DPPC vesicles alone or with calcium resembled those stained with phosphotungstic acid. The electron microscope shows morphologic evidence of fusion but does not allow us to distinguish between the fusion of small vesicles and the somewhat less likely possibility that the small vesicles disintegrate and then form large vesicles. The increased trapped volume (see Table 11) and the conservation of encapsulated carboxfluorescein upon addition of Ca2+ to vesicletubulin complexes (see Fig. 6A), however, indicate that fusion is the most likely mechanism for the formation of the large structures.
Additional evidence for Ca2+-induced fusion of DPPC vesicle-tubulin complexes was provided by the lipid mixing assay developed by Pagano and his coworkers (1 1). Efficient energy transfer was observed between two fluorescent lipid analogues incorporated into the same vesicle containing one of each probe molecule per 100 phospholipid molecules. Upon fusion of these vesicles with a second vesicle population containing no fluorescent lipid, the efficiency of energy transfer was reduced since lateral diffusion of the probes in the plane of the newly formed larger membrane effectively lowers its surface density. In the absence of added Ca2+ the vesicle-tubulin complexes aggregated but did not fuse, and we did not see any changes in energy transfer efficiency. However, upon addition of CaZ+ we observed marked changes in energy transfer. This observation provides a built-in control that aggregation of vesicles does not change the energy transfer pattern. We could not use the resonance energy transfer assay developed by Wilschut et al. (22) based on mixing of intravesicular compartments during fusion; one of the probes used in that assay bound too strongly to tubulin, interfering with this assay.
The energy transfer assay strengthens the interpretation of our data in terms of fusion. Both gel phase and fluid phase (at 44 "C) DPPC vesicles exhibit tubulin-mediated Ca2'-dependent fusion. It it interesting that tubulin, which has been inserted into vesicles, is capable of inducing fusion with pure lipid vesicles. This suggests that the tubulin is altered by its insertion such that the exposed region is more capable of Caz+-dependent perturbation of another lipid bilayer. Such an alteration is consistent with our previous findings of an overall change in the conformation of the molecule upon insertion into the lipid bilayer.
In the majority of earlier studies (22) the effect of divalent cations, mainly Ca", in inducing aggregation and fusion of vesicles has been studied with vesicles made of negatively charged phosphatidylserine or of mixed phospholipids (phosphatidylserine, phosphatidylcholine, and phosphatidylethanolamine). Small unilamellar DPPC vesicles have been reported to fuse (in the absence of Ca2+) to vesicles about 70 nm in diameter, when held below Tc (23). This spontaneous fusion does not produce the large structures seen in Fig. 4 0 . We have earlier reported that the presence of Ca2+ in the medium did not affect the insertion of tubulin into DPPC vesicles at the phase transition (6).
In a recent paper, Hong et al. (24) have presented evidence for the enhancement of Caz+-dependent fusion of vesicles made of phosphatidylserine and phosphatidylethanolamine (1:3) by synexin (M, = 47,000), a water-soluble protein isolated from the adrenal medulla. In another study Zimmerberg et al. (25) observed fusion of phospholipid multilamellar vesicles with a planar phospholipid bilayer membrane that contained a water-soluble Ca2+-binding protein (Mr = 16,000) purified from calf brain. This is, however, the first example of nonleaky fusion of phosphatidylcholine vesicles mediated by a protein and induced by Ca2+.