Evidence that tubulin forms an integral membrane skeleton in molluscan gill cilia

Previous work has demonstrated that scallop ciliary membrane tubulin can be reassociated with natural lipids simply by removal of the solubilizing detergent followed by a freeze-thaw cycle, resulting in monodisperse vesicles with the same protein and lipid composition as the original ciliary membrane (Stephens, 1983). This simple membrane reconstitution may be carried out through at least three additional cycles with little change in composition. Soluble tubulin from the ciliary axoneme is neither incorporated into nor irreversibly adsorbed onto membranes during reconstitution (Stephens, 1985a). Several independent lines of evidence suggest that ciliary membrane tubulin is associated with numerous minor proteins and with lipids, forming a high molecular weight complex (Stephens, 1985a). The lipid may be displaced by condensation with Triton X-114 above the cloud point, forming sedimentable lipid-detergent mixed micelles, while the proteins are rendered soluble by complex formation with detergent, forming particles of uniform sedimentation rate and composition (Stephens, 1985b). Taken together, these data would suggest that membrane tubulin and other proteins form a tightly associated natural complex within the ciliary membrane. Observations of membrane ‘skeletons’ in erythrocytes (Yu et al. 1973; Byers & Branton, 1985) and tumour cells (Ben-Ze’ev et al. 1979; Apgar et al. 1985) after extraction with Triton X-100 prompted us to reinvestigate the nature of tubulin-containing ciliary membranes using gentle detergent extraction. This present electron microscopic study complements and extends earlier work wherein it was shown that Triton X-100 at many times the critical micelle concentration was required to solubilize any appreciable amount of ciliary membrane protein (Stephens, 1985a).

Cilia were prepared from excised gills of the bay scallop (Aequipecten irradia ns) by hypertonic shock followed by differential centrifugation. The membranes were solubilized with 0·5% Triton X-100 in 30 mM-Tris HC1, pH 8·0, 3mM-MgCl₂ and 0·1% mercaptoethanol (EtSH). After centrifugation at 25 000g for 15 min to remove the 9 + 2 axonemes the detergent was removed from the supernatant by adsorption to polystyrene beads (SM-2 Bio Beads) and membranes were reconstituted by freezing and thawing the resultant solution (Stephens, 1983). Reconstituted membranes were recovered by centrifugation at 45 000g for 15 min. Some reconstituted membrane preparations were resolubilized and carried through a second cycle of reconstitution (Stephens, 1985a).

Membrane preparations were resuspended in at least 50 volumes of the appropriate detergent in the above Tris-Mg-EtSH buffer and extracted on ice for 10 min. Whole gill tissue was extracted by gently transferring from salt water into a physiological saline solution developed for reactivation studies (‘extraction solution’ Stommel, 1986), containing the experimental detergent concentration. Peroxide-free Triton X-100 was obtained from Pierce Chemical Company, octyl β-D-glucopyranoside (octyl glucoside; OG) was obtained from Calbiochem-Behring, and Triton X-114 was obtained from Sigma Chemical Company. Analytical ultracentrifugation was performed as previously described (Stephens, 1985a).

Thin pellets (<1 mm) of reconstituted membranes, or several gill filaments, were fixed for 30min with 2·5% glutaraldehyde in the above Tris-Mg buffer, rinsed, and post-fixed for 30–60min with Kar-novsky’s osmium—ferrocyanide (Karnovskv, 1971). The material was stained en bloc overnight with aqueous 1% uranyl acetate, dehydrated with ethanol and embedded in Epon-Araldite resin. Ultrathin sections were observed and photographed with a Zeiss EM-10C electron microscope.

A rabbit polyclonal antibody raised against vinblastine-precipitated sea-urchin egg cytoplasmic tubulin (Fugiwara & Pollard, 1978) was obtained from Polysciences, Inc. (#17870) and a monoclonal antibody against rat brain β-tubulin (Gozes & Barnstable, 1982) was obtained from Miles Laboratories (#63–781). The ability of these antibodies to cross-react with molluscan ciliary membrane and axonemal tubulin subunits was demonstrated by SDS—polyacrylamide gel electrophoresis, as described previously (Stephens, 1985a), followed by immunoblotting to nitrocellulose (Gershoni & Palade, 1982), using alkaline phosphatase-coupled goat anti-rabbit or anti-mouse antibody and colour-development reagents (‘ProtoBlot’) obtained through Promega Biotec. Gill tissue and extracted reconstituted membrane preparations were fixed for 30 min in 2·5% glutaraldehyde and embedded in Lowicryl K4M resin by modifications of the methods of Altman et al. (1984) and Valentino et al. (1985). Thin sections on nickel grids were floated on drops of reagent (one side only; no peroxide etching). They were initially blocked for 5 min with bovine serum albumin at 1 mg ml⁻¹ in Tris-buffered saline (0–15 M-NaCl, 10 mM-Tris HC1, pH 8) containing 0·05% Tween-20, incubated for 30–45 min with rabbit antitubulin at a 1:100 dilution (or rat brain monoclonal at 1:1), rinsed thoroughly with saline, and then incubated for 30—45 min with goat antirabbit (or anti-mouse) IgGs coupled to 5 nm gold particles (E-Y Laboratories) at 1:10 dilution. The sections were thoroughly rinsed and then stained with 2% uranyl acetate in ethanol.

Gill tissue was fixed for 20 min in 2·5% glutaraldehyde in sea water. For cryoprotection, the tissue was equilibrated with 25% glycerol in 0·1 M cacodylate buffer (pH 7·0) that had been adjusted to a tonicity equal to sea water by addition of sucrose. Several filaments were dissected, placed upright on a specimen holder, plunged into liquid Freon 22 cooled in liquid nitrogen, and then frozen in liquid nitrogen. The specimens were mounted in a Balzers Model 300 instrument, cleaved across the filament axis at —115°C and 0^·5 × 10⁻⁸ Torr, and etched for 1 min at — 100°C. Platinum/carbon-shadowed replicas were examined as above.

By resuspending reconstituted membranes in increasing concentrations of detergent, one can observe critical points in the solubilization process. At concentrations of NP-40 below the critical micelle concentration, no disruption is evident and, in fact, incompletely resealed membrane leaflets appear to coalesce to form more uniform vesicles, since they are far less evident in detergent-treated material (not shown). It was shown elsewhere that vesicle preparations cleared optically at the critical micelle concentration (Stephens, 1985a) but it is obvious from Fig. 1A-B that the structure of the membrane remains basically intact. Under even higher detergent conditions, most of the membrane protein in either reconstituted vesicles or isolated cilia is still sedimentable at low speed (Stephens, 1985a). Fig. 1C demonstrates that this is so because a skeleton of the membrane remains. The same results are obtained with highly purified Triton X-100, anti-oxidant treated Triton X114, or octyl glucoside, in the presence or absence of mercaptoethanol, eliminating -SH oxidation and crosslinkage as a cause for the skeletal structure. At >5 × critical micelle concentration of NP-40 (or at slightly greater than the critical micelle concentration of octyl glucoside), solubilization takes place, resulting in formation of a granulo-reticular material with a very low sedimentation coefficient (see below).

One can take advantage of the ability to reconstitute membranes from octyl glucoside solutions by dilution to show that the solubilization to a granulo-reticular material is fully reversible. At a concentration of octyl glucoside slightly greater than the critical micelle concentration, the membrane proteins are fully solubilized, yielding a 4S protein-lipid-detergent mixed micelle (Stephens, 1985<7). Negative staining or sedimentation and fixation of similar Triton-solubilized micellar material yields 5–10 nm particles (minimum and maximum estimate; average 6·5±0·8nm) and aggregates thereof (Stephens, 1985a,b), as does freeze-thaw concentration of the octyl glucoside-solubilized material (Fig. 2B). When such a solution is diluted back to the critical micelle concentration, membranes begin to reform in conjunction with the granulo-reticular material (Fig. 2C). Further dilution to 1/5 the critical micelle concentration results in formation of a relatively uniform population of vesicles (Fig. 2D), closely resembling the twice-recycled vesicle preparation from which the solubilized protein was derived (Fig. 2A). Although such a result does not prove that the membrane tubulin is native, this reversibility upon dilution demonstrates that the granulo-reticular material consistently obtained when total solubilization is achieved is not simply denatured protein since it totally vanishes as membranes reform.

Previous work suggests that the particulate appearance of the detergent-solubilized ciliary membrane most likely reflects the formation of large protein—detergent mixed micelles (Stephens, 1985a,b) but the question remains whether comparably large protein complexes exist naturally in the ciliary membrane. Consequently, we investigated ciliary membranes in situ by freezefracture. Fig. 3 illustrates a representative view of ciliary axonemal membranes (A), the proximal or basal regions of the cilia (B), and the junctional region between two lateral cells (C). Although the characteristic 3-stranded ciliary necklaces and also the particle arrays associated with basal outer doublet-membrane linkages are both clearly evident (cf. Dute & Kung, 1978), few particles are present in the membrane of the ciliary axoneme proper. If the ‘particles’ or aggregates seen after detergent removal represented natural protein complexes in the axonemal membrane, one might expect to see numerous intramembrane particles of a size comparable to those of the ciliary necklace (7–8nm). The ciliary shaft membranes are, for the most part, smooth. What few particles do exist are sparse and show no linear organization along the underlying outer doublets, contrary to what might be expected for this very protein-rich membrane with extensive microtubule-membrane interactions (cf. Dentier et al. 1980). Both septate and gap junction particle arrays are well-preserved (C), indicating optimal fixation. (Gap junctions in lamellibranch molluscs can be dissociated or dispersed under certain ionic conditions and we have used their appearance as a criterion to assess preservation (cf. Good & Stephens, 1987).)

Using the same conditions under which membrane skeletons are obtained from reconstituted vesicles, one can examine the membrane structure of gill cilia in situ. When gill tissue is extracted with 0·06% NP-40, a membrane ‘sleeve’ remains around the cilia (Fig. 4A). At high magnification this sleeve is seen to be attached to the axoneme by fine bridges where the section is precisely through and parallel to an outer doublet (Fig. 4B). These bridges have an occasional periodicity of 20–24 nm (cf. Dentier et al. 1980). At this detergent concentration, vesicular material appears to ‘bud-off’ the membrane, probably representing lipid-detergent micelles, since no appreciable membrane protein is solubilized (cf. Stommel et al. 1982). Beyond this detergent concentration, the membrane remnant simply disappears as fragments into the medium. When isolated cilia are treated under comparable conditions, by resuspension in 0·06% detergent followed by sedimentation, very little membrane protein solubilization results (cf. Stephens, 1985a). In this case, the remnant membrane material is stripped from the axoneme by repeated pipetting during extraction and by shear during centrifugation, appearing as a granulo-reticular network that co-sediments with the axonemes (Fig. 4C).

To determine whether antigenic sites of membrane tubulin are exposed at the membrane surface, either as a consequence of natural membrane topology or as a result of artifactual adsorption of tubulin to the membrane, we attempted to detect membrane tubulin by immunogold labelling. The polyclonal antibody that we used was raised against sea-urchin egg cytoplasmic tubulin but we found that it cross-reacts well with the two tubulin subunits of both ciliary axonemes and recycled, reconstituted membranes at essentially equivalent titres (Fig. 5A-B). In addition, parallel studies were carried out with a monoclonal anti-rat brain β-tubulin that has comparable cross-reactivity toward membrane and axonemal β-tubulin chains (Fig. 5C).

In Lowicryl K4M thin sections of ciliary axonemes, used here as an internal control, these antibodies are readily localized by 5-nm gold-labelled anti-rabbit and anti-mouse IgGs, staining the outer doublet and central pair microtubules (Fig. 6A). With the important qualification that the size of the (double) antibodies and the gold particles restrict resolution, we have not been able to demonstrate, unequivocally, an additional ring of labelling around the periphery of the axoneme, bevond the doublets, as would be expected if the membrane-associated tubulin were also crossreactive. However, this point is complicated by the fact that the membrane is more closclv apposed to the axoneme when osmium fixation, which destroys antigenicity, is omitted (W. L. Dentier, personal communication).

We are also unable to detect unequivocal gold labelling of untreated reconstituted membrane vesicles but this observation is complicated by poor preservation of closed vesicles in the absence of osmium postfixation (not shown). In contrast, partially extracted vesicles are well-preserved, and when a preparation containing both membranous and membrane skeletal material (comparable to that in Fig. 1C) is reacted with sea-urchin anti-tubulin and gold-labelled anti-IgG, most of the labelling occurs on the membrane skeletal material (Fig. 6B). The near-absence of label on intact membrane leaflets (arrows) would argue against arti-factually-absorbed tubulin, while the irregular labelling of the partially-extracted reticular skeletal material would suggest that the antigenic sites of tubulin might be masked in regions higher in lipid. No differences in localization are observed when labelling by the primarily anti-α tubulin polyclonal antibody is compared with that of the β tubulin monoclonal, except that the former results in higher overall labelling, an obvious reflection of its less stringent specificity.

The evidence presented here suggests that membrane tubulin and its associated proteins constitute an integral protein skeleton in the ciliary membrane since gentle delipidation leaves behind most membrane proteins in the form of a ‘ghost’ or ‘membrane matrix’ whose topology is essentially identical to that of the original ciliary membrane or reconstituted vesicles. Earlier work showed that there is no selective extraction of any major membrane protein as detergent concentration is increased; rather, when the detergent to lipid weight ratio reaches >5 :1, all proteins become soluble (Stephens, 1985a). Although high detergent concentrations produce a granulo-reticular material, this apparently is the result of micellization, since freeze-fracture reveals no significant number of membrane particles of the expected size in the ciliary membrane proper.

Considering that about 2/3 of the total ciliary membrane protein is tubulin, it is surprising that no unequivocal antibody labelling was detected on intact membranes. It would appear that both of the membrane tubulin polypeptide chains are relatively inaccessible, at least to the antibodies that we tested, although the possibility still remains that different antigenic determinants may be exposed at the membrane surface or that preparation for electron microscopy selectively leaves only internal antigenic sites intact. Alternatively, in the case of partially extracted membrane vesicles, the near-absence of antibody labelling on lamellar membranes may indicate that these arc pure lipid, a consequence of membrane proteins having gathered into the granulo-reticular material that docs cross-react.

A clarification of nomenclature is important at this point since workers in the membrane structure field use ‘membrane matrix’ in the sense of that which is left when a membrane is extracted by detergent. This is in contrast to usage in the cilia field where, for decades, this term has referred to a fraction obtained by extraction of isolated cilia with high concentrations of detergent. This fraction contains all of the membrane proteins plus any soluble periaxonemal proteins, i.e. everything but the 9 + 2 axoneme structure itself. Interestingly, at least in gill cilia, the bulk of this membrane plus periaxonemal matrix fraction is in fact the membrane matrix (i.e. skeleton) proteins, as reassociation studies have shown. The only major exception is calmodulin, a large fraction of which is free within the periaxonemal space (Stommel et al. 1982; Stephens, 1983).

Our present results support conclusions about ciliary membrane protein topology that we reached earlier from lipophilic cross-linkage studies of isolated cilia from scallop and Tetrahymena (Dentier et al. 1980). Some of the membrane tubulin, a sub-set of other membrane proteins, and a high molecular weight dynein-like polypeptide were judged to be in proximity, corresponding to a stable outer doublet-membrane connection whose morphology resembles the fine bridges that we see here after gentle detergent extraction of intact cilia. In addition, the remaining membrane tubulin was judged to be integral since it became cross-linked to itself, either as dimers or very high molecular weight material, through the action of the photoactivated lipophilic agent. We concluded that dynein-like bridges provided a tethering mechanism by which the tubulin-containing ciliary membrane was held to the vigorously beating cilium. Using the very different approach of direct antibody labelling, Mar-chese-Ragona & Johnson (1985) report that 14 S dynein is located between the outer doublets and the adjacent membrane in Tetrahymena cilia, whereas 30 S dynein corresponds to the classic arms.

These results do not shed any further light on the actual function of ciliary membrane tubulin but they are at least consistent with one function postulated earlier, that of signal transduction (Stephens, 1986). Membrane tubulin has been reported mainlv in cilia that respond to touch. The strong association among tubulin and other membrane proteins into a membrane skeleton, tethered to the axoneme, could provide rigidity toward mechanical deformation in these mechanosensitivc cilia, either resulting in transduction within the ciliary membrane itself or transmission of force to the basal region or somatic membrane. Our earlier work showed that increasing the rigidity of the ciliary membrane of molluscan statocyst sensory cilia with a lipophilic cross-linker markedly enhanced transmission of applied force and amplified the resulting electrical response (Stommel et al. 1980).

This work was supported by USP11S Grants GM 20 644 and GM 29 503 from the National Institute for General Medical Sciences. We thank Robert Crowther and Louis Kerr for untiring assistance with imniunocyto-chemistry and freeze-fracture electron microscopy, respectively.