Tubulin, Gq, and phosphatidylinositol 4,5-bisphosphate interact to regulate phospholipase Cbeta1 signaling.

The cytoskeletal protein, tubulin, has been shown to regulate adenylyl cyclase activity through its interaction with the specific G protein α subunits, Gαs or Gαi1. Tubulin activates these G proteins by transferring GTP and stabilizing the active nucleotide-bound Gα conformation. To study the possibility of tubulin involvement in Gαq-mediated phospholipase Cβ1 (PLCβ1) signaling, the m1 muscarinic receptor, Gαq, and PLCβ1 were expressed in Sf9 cells. A unique ability of tubulin to regulate PLCβ1 was observed. Low concentrations of tubulin, with guanine nucleotide bound, activated PLCβ1, whereas higher concentrations inhibited the enzyme. Interaction of tubulin with both Gαq and PLCβ1, accompanied by guanine nucleotide transfer from tubulin to Gαq, is suggested as a mechanism for the enzyme activation. The PLCβ1 substrate, phosphatidylinositol 4,5-bisphosphate, bound to tubulin and prevented microtubule assembly. This observation suggested a mechanism for the inhibition of PLCβ1 by tubulin, since high tubulin concentrations might prevent the access of PLCβ1 to its substrate. Activation of m1 muscarinic receptors by carbachol relaxed this inhibition, probably by increasing the affinity of Gαq for tubulin. Involvement of tubulin in the articulation between PLCβ1 signaling and microtubule assembly might prove important for the intracellular governing of a broad range of cellular events.

The cytoskeletal GTP-binding protein tubulin activates or inhibits adenylyl cyclase through its interaction with G␣ s or G␣ i1 (13)(14)(15)(16). Tubulin binds to these proteins and activates them via direct transfer of GTP. Both guanine nucleotide transfer and the stabilization of the active G␣ conformation are thought responsible for the sustained G protein activation by tubulin (15,16). Complexes of dimeric tubulin with G␣ s and G␣ i1 exist in the synaptic membrane (17), and these complexes provide the physical framework for the interface between G protein-mediated signal transduction and the cytoskeleton. Previous work has focused on the regulation of adenylyl cyclase by tubulin, but the possibilities that other G protein-mediated signaling enzymes suffered similar regulation waranted investigation.
Although an association of PLC with the turkey erythrocyte cytoskeleton has been reported (18), the effect of tubulin on PLC␤ 1 signaling has not been examined. The current study was designed to determine whether tubulin plays a role in the regulation of PLC␤ 1 , and, conversely, how phosphoinositides might affect the ability of tubulin to assemble into microtubules or to interact with G proteins. This was approached using Sf9 cells expressing various combinations of m 1 muscarinic receptors, G␣ q , and PLC␤ 1 . It was determined that tubulin does indeed bind to G␣ q and activates that molecule via the direct transfer of GTP. It is also demonstrated that PLC␤ 1 may enjoy a dual regulation by dimeric tubulin, the molecule inhibiting PLC␤ 1 by binding to the enzyme substrate PIP 2 . This represents another arena in which cell structure and cell signaling share a complex regulatory liaison.
Receptor binding studies using [ 3 H]QNB (0.02-4.00 nM) as a ligand were also performed to assess the number and affinity of the m 1 muscarinic receptors (23). The nonspecific binding, measured in the presence of 1 M atropine, was less than 10%. Saturation isoterms were analyzed using the LUNDON 1 program.
Purification of PLC␤ 1 -cDNA encoding the entire rat PLC␤ 1 amino acid sequence was cloned into the pSC11 vaccinia virus expression vector and HeLa cells were infected with recombinant virus as described (24). The enzyme was extracted from the membrane with 2 M KCl for 2 h. The extracted PLC␤ 1 was consecutively subjected to preparative HPLC on phenyl-5PW (150 ϫ 21.5-mm column), analytical HPLC on heparin-5PW (75 ϫ 7.5-mm column), and ion-exchange HPLC on Mono-Q FPLC (60 ϫ 7-mm column) chromatography as described (24).
Tubulin Preparations-Microtubule proteins were prepared as described (25). Microtubule-associated proteins were removed by phosphocellulose chromatography (PC-tubulin) as described (26). Tubulin-GppNHp or tubulin-[ 32 P]AAGTP were prepared from PC-tubulin by the removal of GTP by means of charcoal pretreatment, followed by incubation in the presence of 150 M of GppNHp (or [ 32 P]AAGTP) for 30 min on ice as described (27). Prior to use, tubulin-guanine nucleotide preparations were passed through P6-DG columns twice in order to remove unbound nucleotide. After this procedure 0.4 -0.6 mol of nucleotide were bound per mol of tubulin. This remained stable even after five desalting steps (27). Tubulin-guanine nucleotide concentrations used throughout the study were based on the protein concentration. [ 32 P]AAGTP and AAGTP were synthesized as described (28).  (29) and deoxycholate to give a final concentration of 1 mM) were incubated in a Branson water bath sonicator for 15 min at 4°C. Ten l of GppNHp, tubulin-GppNHp, or tubulin, with or without 10 l of carbachol, were added at appropriate concentrations to a final volume of 80 l, and the tubes were incubated for 15 min at 37°C with constant shaking. Aqueous and lipid phases were separated as described (30)

, and [ 3 H]inositol trisphosphate ([ 3 H]IP 3 ) in the aqueous phase was quantified by liquid scintillation counting.
Reconstitution of Recombinant G␣ q and PLC␤ 1 -Sf9 cells were infected with G␣ q baculovirus and extracted after nitrogen cavitation (500 p.s.i. for 20 min) with 0.1% CHAPS in 20 mM Hepes, pH 7.5, 1 mM EDTA, 3 mM MgCl 2 , 150 mM NaCl, 0.1 mM phenylmethylsulfonyl fluoride, and 2 g/ml each of aprotinin, leupeptin, pepstatin, and benzamidine at 4°C for 1 h with constant stirring. The tubes were centrifuged at 100,000 ϫ g for 30 min at 4°C. G␣ q -containing extracts (protein concentrations as indicated) were reconstituted into lipid vesicles (phosphatidylethanolamine:PIP 2 at a 4:1 molar ratio) by means of sonication in 20 mM Tris maleate, pH 6.8, 6 mM MgCl 2 , 30 mM NaCl, 120 mM KCl, and EGTA and CaCl 2 to give a final concentration of 50 nM Ca 2ϩ (29). The PIP 2 final concentration was 30 M (1 Ci of [ 3 H]PIP 2 ). 40 l of this mixture was transferred to each experimental tube; 20 l of purified recombinant PLC␤ 1 (protein concentration as indicated) in 100 g/ml bovine serum albumin was added, and the tubes were sonicated in a Branson water bath sonicator for 15 min at 4°C as described above. Ten l of GppNHp, tubulin-GppNHp, or tubulin, stripped of nucleotide, was added at appropriate concentrations to a final volume of 90 l. The samples were incubated for 10 min at 30°C with constant shaking. The reaction was stopped and [ 3 H]IP 3 was quantified as described above.
Photoaffinity Labeling and Nucleotide Transfer-Sf9 cell membranes, expressing recombinant proteins, were incubated with 1 M tubulin-[ 32 P]AAGTP in the presence or absence of 1 mM carbachol in 100 mM Pipes buffer, pH 6.9, 2 mM EGTA, 1 mM MgCl 2 (buffer A) for 10 min at 23°C with constant shaking. The tubes were UV-irradiated for 4 min on ice, and the reaction was quenched with ice-cold Hank's buffer, 1 mM MgCl 2 , 4 mM DTT. After centrifugation at 100,000 ϫ g for 15 min, the membrane pellets and concentrated supernatants (Amicon, Centriprep 10) were dissolved in 3% SDS Laemmli sample buffer with 50 mM DTT. Membrane (70 g of protein) and supernatant fractions (20 g of protein) were subjected to SDS-PAGE (10% acrylamide and 0.133% bisacrylamide) as described (31). Gels were either stained (Coomassie Blue) or subjected to Western blotting, followed by autoradiography (Kodak XAR-5 film). The radioactivity of the bands was measured with a PhosphorImager (Molecular Dynamics).
Immunoprecipitation-Sf9 cells were infected separately with baculoviruses bearing either the m 1 muscarinic receptor, G␣ q , or PLC␤ 1 cDNA. Each of the three different membrane preparations was extracted with 1% sodium cholate in buffer A for 1 h at 4°C with constant stirring. The tubes were centrifuged at 20,000 ϫ g for 15 min at 4°C. Membrane extracts (0.5 mg/ml membrane protein) were incubated with 1 M tubulin-[ 32 P]AAGTP, as described above. After UV irradiation and preclearing (Calbiochem Standardized Pansorbin cells, manufacturer's instruction), each membrane extract was incubated overnight with appropriate specific or nonspecific antiserum (1:20 dilution) at 4°C with constant shaking. Immune complexes were precipitated with pansorbin cells, and 100 g of protein of each immunoprecipitate were subjected to SDS-PAGE and autoradiography. The antisera used showed no cross-reactivity to tubulin.
Gel Filtration Assay (32)-Homogeneous PIP 2 micelles were prepared by suspending 100 g of PIP 2 in 0.5 ml of buffer A and sonicating on ice for 5 min. Tubulin was incubated with PIP 2 micelles in a Branson water bath sonicator for 15 min at 4°C, followed by an additional 45 min on ice. The samples were run on a 0.7 ϫ 50-cm column of Ultrogel AcA 34 at 4°C. The column was equilibrated with buffer A (flow rate ϭ 20 ml/h), and fractions of 0.5 ml were collected and assayed for protein.
The data are given in arbitrary units since lipids quench the Bradford dye-binding assay (32).
Electron Microscopy-Samples were taken from tubulin polymerization reactions (2 mg/ml final protein concentration) carried out in buffer A, containing 3 mM MgCl 2 , 1 mM GTP, and 30% glycerol, for 30 min at 37°C with constant shaking (26). PIP 2 , when added, was sonicated in buffer A as described. 10 l of the reaction mixture was applied to a carbon/Formvar-coated grid for about 10 s. The grids were stained with several drops of 1% uranyl acetate and air-dried. Observations were made in a JEM 100CX (JEOL) electron microscope. For microtubule length determinations, at least 100 assembled structures were measured in multiple fields from two different grids representing two different experiments.

RESULTS AND DISCUSSION
Receptor-independent Regulation of PLC␤ 1 by Tubulin-Simultaneous baculovirus-mediated expression of G␣ q and PLC␤ 1 was performed in Sf9 cells, and the effects of GppNHp and tubulin with GppNHp bound (tubulin-GppNHp) on G␣ qregulated PLC␤ 1 activity were studied in membranes prepared from infected cells (Fig. 1). GppNHp increased the already high basal PLC␤ 1 activity in a concentration-dependent and saturable manner (Fig. 1A), confirming the functional coupling between the recombinant G␣ q and PLC␤ 1 . GppNHp (at 100 M) did not affect PLC␤ 1 activity in membranes from cells expressing only PLC␤ 1 (Table I), thus indicating that neither G␤␥ nor any G␣ resident in Sf9 cell membranes is capable of activating the expressed PLC␤ 1 . Expressed G␣ q did not couple to any endogenous PLC activity. Endogenous PLC activity was also undetectable in the presence or absence of GppNHp. Tubulin-GppNHp had been shown to activate adenylyl cyclase in membranes from either COS 1 or C6 glioma cells (16,17) in a manner similar to GppNHp. Surprisingly, it appeared that in membranes prepared from Sf9 cells expressing G␣ q and PLC␤ 1 tubulin-GppNHp inhibited PLC␤ 1 under the same conditions that GppNHp stimulated the enzyme (Fig. 1B).
To study this further, extracts of Sf9 cells expressing G␣ q were reconstituted with purified recombinant PLC␤ 1 and exogenous phospholipid vesicles and assayed under conditions similar to those used for Sf9 cell membranes (Fig. 2). The effects of tubulin-GppNHp, nucleotide-free tubulin, and GppNHp were compared. Although GppNHp stimulated PLC␤ 1 activity at concentrations higher than 1 M, 30 nM tubulin-GppNHp activated the enzyme. However, at higher tubulin-GppNHp concentrations inhibition of PLC␤ 1 was observed. Nucleotide-free tubulin did not activate but only inhibited PLC␤ 1 at concentrations higher than 300 nM. Tubulin Modulation of m 1 Muscarinic Receptor Activation of PLC␤ 1 -Involvement of tubulin in receptor-triggered PLC␤ 1 regulation was studied with membranes from Sf9 cells expressing m 1 muscarinic receptors, G␣ q , and PLC␤ 1 . The m 1 muscarinic receptor expression level was estimated by receptor binding studies with [ 3 H]QNB as a ligand. When all three recombinant proteins were expressed, the m 1 muscarinic receptor binding capacity (B max ) was 240 Ϯ 21 fmol/mg mem-brane protein and K d ϭ 0.162 Ϯ 0.010 nM (n ϭ 3). The ability to construct a complete, receptor-activated and G protein-mediated system allowed a comparison of the effects of GppNHp, tubulin-GppNHp, and tubulin, denuded of exchangeable nucleotide, on carbachol-stimulated PLC␤ 1 activity (Fig. 3). Although GppNHp was able to activate PLC␤ 1 without receptor stimulation, the effect was potentiated by carbachol (Fig. 3A). The response of PLC␤ 1 to GppNHp (up to 1 mM) under these conditions was concentration-dependent, saturable, and potentiated by carbachol throughout the concentration range (50.5 Ϯ 9.2% at 100 M GppNHp). However, the effect of tubulin-GppNHp on PLC␤ 1 activity was again biphasic, stimulatory at the lower (30 nM) and inhibitory at the higher concentrations of tubulin (Fig. 3A). Carbachol potentiated the tubulin-GppNHpevoked PLC␤ 1 activation. At 30 nM, tubulin-GppNHp was more efficacious than GppNHp, independent of the addition of carbachol. The effect of carbachol was receptor-mediated since, at tubulin concentrations (30 nM) that stimulated PLC␤ 1 , atropine inhibited carbachol-induced PLC␤ 1 activity by about 80% (Fig. 3B). However, the m 1 muscarinic receptor stimulation was able to overcome the inhibitory effect of higher tubulin concentrations, resulting in an effective PLC␤ 1 activation. Note   that the highest tubulin concentrations used were below those where tubulin-GppNHp forms polymers. To clarify the mechanism of this dual regulation of PLC␤ 1 by tubulin, the effect of nucleotide-free tubulin on enzyme activity was studied (Fig.  3C). Tubulin (without nucleotide) inhibited PLC␤ 1 activity in a concentration-dependent manner. No enzyme activation was observed at any tested tubulin concentration. Furthermore, unlike tubulin-GppNHp, nucleotide-free tubulin did not potentiate the PLC␤ 1 activation induced by carbachol. Tubulin-GDP or tubulin-GDP␤S (at 30 nM) also failed to potentiate carbacholtriggered PLC␤ 1 activation or to affect the basal enzyme activity. Thus, it appears that tubulin activates PLC␤ 1 only when GTP or GTP analog occupies the exchangeable GTP-binding site. 2 Activation of G␣ q by Tubulin and Recruitment of Tubulin to the Plasma Membrane-To study directly the ability of tubulinguanine nucleotide to interact with G␣ q , the transfer of the hydrolysis-resistant photoaffinity GTP analog, [ 32 P]AAGTP, from tubulin to G␣ q , was examined in Sf9 cells expressing the m 1 muscarinic receptor, G␣ q and PLC␤ 1 . Nucleotide transfer from tubulin appears to be responsible for the activation of G␣ s and G␣ i1 , and this has been observed in permeable cells, membrane preparations, and reconstituted systems (13)(14)(15)(16)(17). It has been shown that G␣ proteins receive [ 32 P]AAGTP from tubulin under conditions when G␣ is incapable of binding the free nucleotide. This appears to be due to the formation of a complex between tubulin and G␣. Under these conditions, tubulin retains the nucleotide without releasing it to the media (15). Carbachol (1 mM) increased (36 Ϯ 7%) the binding of tubulin-[ 32 P]AAGTP to the membrane, and there was commensurate loss (45 Ϯ 8%) of tubulin-[ 32 P]AAGTP from the supernatant (Fig. 4). Transfer of [ 32 P]AAGTP from tubulin to G␣ q was also potentiated (42 Ϯ 9%) by carbachol. Atropine blocked carbachol-induced binding of tubulin to the membrane as well as the increase in [ 32 P]AAGTP transfer. 3 The manner in which the activated m 1 muscarinic receptor increases tubulin association with the membrane to regulate PLC␤ 1 is currently under study. The increase in tubulin-GppNHp activation of PLC␤ 1 brought by carbachol may be explained partially by this facilitated association of tubulin with the membrane.
These findings, together with the results in Figs. 2 and 3, support the idea that nucleotide transfer from tubulin to G␣ q is a potential mechanism for PLC␤ 1 activation. Carbachol had no effect on the membrane association of tubulin, nucleotide transfer to G␣ q , or PLC␤ 1 activity in Sf9 cells expressing only G␣ q and PLC␤ 1 . No effect of carbachol, guanine nucleotide, or tubulin-guanine nucleotide was observed in control Sf9 cells when measuring PIP 2 hydrolysis or photoaffinity labeling.
Tubulin, G␣ q , and PLC␤ 1 Form a Complex-To understand better the mode of interaction of tubulin-GppNHp with the members of PLC␤ 1 signaling cascade, coimmunoprecipitation studies with tubulin-[ 32 P]AAGTP were performed. As seen in   3. Effect of m 1 muscarinic receptor stimulation on PLC␤ 1 activity regulated by tubulin. Sf9 cells expressing recombinant m 1 muscarinic receptors, G␣ q , and PLC␤ 1 were harvested 65 h after infection, and membranes were prepared as described under "Experimental Procedures." PLC␤ 1 activity was assayed using 20 g of membrane protein. A, effect of carbachol (carb) on GppNHp or tubulin-GppNHp (Tub-GppNHp) regulated PLC␤ 1 activity. Indicated concentrations of carbachol, GppNHp, or tubulin-GppNHp were added to the membranes, and PLC␤ 1 activity was assayed as described above. PLC␤ 1 activity was 1.11 Ϯ 0.03 nmol/min/mg protein in the presence of carbachol alone. B, atropine inhibition of carbachol-induced activation of PLC␤ 1 in the presence of tubulin-GppNHp. PLC␤ 1 activity was studied in the pres- Fig. 5, both G␣ q antiserum and, to a lesser extent, PLC␤ 1 antiserum coimmunoprecipitated tubulin when membrane extracts of cells infected with viruses for G␣ q or PLC␤ 1 , respectively, were studied. In the membrane extracts from cells expressing G␣ q , [ 32 P]AAGTP-labeling of G␣ q was also observed. The source of [ 32 P]AAGTP was tubulin. No tubulin-[ 32 P]AAGTP coimmunoprecipitation was observed when extracts from normal Sf9 cells were tested with anti-G␣ q or anti-PLC␤ 1 antisera. Thus, it is suggested that the formation of a complex between G␣ q , PLC␤ 1 , and tubulin-GppNHp might be responsible for the higher efficacy of tubulin-GppNHp, compared with GppNHp, during the m 1 muscarinic receptor stimulation.
PIP 2 Binding to Tubulin-In order to clarify the inhibitory action of tubulin on PLC␤ 1 , the interaction between tubulin and the PLC␤ 1 substrate, PIP 2 , was studied. It has been shown previously that phosphatidylinositol inhibits microtubule assembly (34) and that the binding of profilin (an actin-binding protein) to PIP 2 micelles inhibits PLC␥ 1 -directed PIP 2 hydrolysis (35). To test the possibility that tubulin binds PIP 2 , a gel filtration assay was performed. Two peaks of tubulin were resolved, one represented a small portion of oligomeric tubulin, and the second represented dimeric tubulin (Fig. 6). When the same amount of tubulin was incubated with PIP 2 micelles and passed over an Ultrogel 34 column, an increase in the higher molecular weight peak was observed along with a corresponding decrease in the dimeric tubulin peak. When tested under the same conditions, phosphatidylcholine micelles did not shift the mobility of tubulin. Since PIP 2 forms micelles of about 93 kDa in aqueous solutions (36), the increase in the higher molecular weight fraction could be attributed to either PIP 2 -induced tubulin oligomerization or to a binding of tubulin to PIP 2 micelles, as has been observed in the case of profilin (32). To resolve this, polymerization of tubulin in the absence (Fig. 7A) or presence (Fig. 7B) of PIP 2 micelles was studied by electron microscopy. Tubulin failed to polymerize properly in the presence of PIP 2 (Fig. 7B). Microtubules were considerably shorter (1.7 Ϯ 1.0 M, compared with 10.2 Ϯ 4.8 M in the absence of PIP 2 ) and 3-4 times thicker (75-100 nm versus 25 nm in the absence of PIP 2 ). These results suggest a direct interaction between tubulin and PIP 2 and support the hypothesis that increased concentrations of dimeric tubulin might prevent access of PLC␤ 1 to its substrate. Since increased activation of PLC␤1 might evoke localized increases in calcium which could increase tubulin dimer concentration, this might represent a feedback mechanism for the regulation of receptor-G proteinactivated phospholipase C. It is noteworthy in this regard that phosphorylation of PLC␥ 1 by EGF receptor tyrosine kinase appears to overcome the inhibitory effect of profilin binding to PIP 2 , resulting in an effective activation of the enzyme (35). Similarly, m 1 muscarinic receptor stimulation might reverse tubulin-evoked inhibition of PLC␤ 1 by increasing the affinity of Membrane preparations of Sf9 cells, expressing only m 1 muscarinic receptors, G␣ q , or PLC␤ 1 were extracted with 1% sodium cholate. Membrane extracts (0.5 mg/ml membrane protein) were incubated with 1 M tubulin-[ 32 P]AAGTP, as described under "Experimental Procedures." After UV irradiation each membrane extract was incubated overnight with appropriate specific or nonspecific antiserum as indicated. Immunoprecipitates were subjected to SDS-PAGE and autoradiography. An autoradiogram from one of four independent experiments with identical results is shown. Tub, tubulin.
FIG. 6. Interaction between tubulin and PIP 2 . Gel filtration assay for the binding of tubulin to PIP 2 micelles was performed. Tubulin was incubated with PIP 2 micelles in a Branson water bath sonicator for 15 min at 4°C, followed by an additional 45 min on ice. The samples were then applied to and eluted from a 0.7 ϫ 50 cm column of Ultrogel AcA 34 at 4°C as described under "Experimental Procedures." Chromatograms of 0.7 mg of tubulin alone (open circles) or with PIP 2 at a molar ratio of 1:14 (closed circles) are shown. When chromatographed alone, PIP 2 micelles at the same concentration did not show detectable 280 nm absorbance in any fraction collected. G␣ q for tubulin.
Feedback Inhibition of PLC␤ 1 by Tubulin-It is suggested that the amount of dimeric tubulin normally accessible to the membrane is low; thus, initially, stimulation of G␣ q by tubulin would predominate. It is possible, however, that tubulin bound to PLC␤ 1 and/or PIP 2 might not bind to G␣ q . The observed muscarinic receptor-triggered increase in G␣ q affinity for tubulin might be able to overcome such inhibition. It is also possible that the down-regulation of PLC␤ 1 signaling would channel heterotrimeric G q to another signaling pathway (37,38). Although complexes of tubulin with G␣ q , similar to those with G␣ i1 or G␣ s (17), have not yet been shown in the membrane, clearly the possibility exists for a system in which membrane-associated tubulin, through reversible interaction with other membrane proteins and/or lipids, could achieve a dual regulation of phospholipase C signaling within the cell. It has also been suggested that the activation of G protein ␣ subunits by tubulin dimer may provide an interface between G protein signaling through cAMP and G protein signaling through calcium (39). These dynamic interactions between G proteins and tubulin dimers at the synapse may also modify synaptic microarchitecture.
A noteworthy feature of this study is that tubulin regulates PLC␤ 1 signaling while the PLC␤ 1 substrate, PIP 2 , may regu-late microtubule assembly. These reciprocal events could be involved both in intracellular signaling and the control of spindle morphogenesis and reorganization of microtubule arrays during the cell cycle. Since alterations in the metabolism of phosphoinositides (40,41) or the expression of m 1 muscarinic receptors (42) or G proteins (43,44) have been implicated in cellular transformation, this regulatory mechanism might prove valuable in the control of cell proliferation as well. FIG. 7. Effect of PIP 2 on microtubule polymerization. Electron microscopy of polymers formed by PC-tubulin in the absence (A) and presence (B) of PIP 2 . Samples were taken from tubulin polymerization reactions (2 mg/ml final protein concentration) carried out for 30 min at 37°C as described under "Experimental Procedures." The molar ratio between tubulin and PIP 2 , when added, was 1:6. Reaction mixtures were applied to carbon/Formvar-coated grids and prepared for electron microscopy as described under "Experimental Procedures." Grids were examined on a JEM 100CX (JEOL) electron microscope. Magnification ϭ 7000 ϫ.