Retention of ciliary ninefold structure after removal of microtubules

The 9+2 structure of cilia and flagella is nearly invariant throughout nature and is the result of a precise interaction among the outer doublet microtubules and the myriad of architectural proteins that make up the interdoublet linkages, radial spokes, and central pair complex. Associated within the outer doublet microtubules is a distinctive class of proteins, the tektins (Linck, 1982), which comprise the insoluble filaments that remain when tubulin is extracted from outer doublets with urea and the detergent Sarkosyl (Linck, 1976; Linck & Langevin, 1982; Linck & Stephens, 1987). These protein filaments are tightly intercalated within the microtubule substructure (Linck et al. 1985) and are thought to be associated with the A –B junctional protofilaments in the region where various architectural proteins attach (Amos et al. 1986). The outer doublet microtubules of cilia are bound together in ninefold symmetry by the circumferential interdoublet, or nexin, linkages that must tie the A-tubules together, since solubilization of the B-tubules by low ionic strength dialysis leaves a cylinder of nine singlet A-tubules (Linck, 1973b). Furthermore, much of the radial spoke structure and one central pair member also remain, suggesting an intimate association of these with the A-tubule wall. Nexin linkages are thought to restrain the sliding of doublet microtubules, converting this linear motion into bending (Summers & Gibbons, 1971), while cyclic interaction of radial spokes with the central pair is thought to be involved in the coordination of sliding (Warner & Satir, 1974).

Our studies are directed towards determining how these various components are associated with microtubules to form the three-dimensional structure of cilia, with particular emphasis on the location of tektins and the attachment of accessory proteins in axonemal microtubules. We show that much of the 9+2 microtubules of scallop gill and sea-urchin embryonic cilia can be melted away, leaving a stable ninefold symmetric remnant of constant composition, containing the tektins and most of the other architectural proteins of the axoneme. A preliminary report of this work has been presented in abstract form (Stephens et al. 1987).

Cilia were prepared from thoroughly washed gills of the bay scallop Aequipecten irradians by hypertonic shock at 20 °C, followed by two cycles of differential centrifugation from 4°C sea-water (Linck, 1973a; Stephens, 1983). Cilia were similarly prepared from sea-urchin embryos (blastula through late gastrula) of Strongylocentmtus droebachiensis (Maine) and Tripneustes gratilla (Hawaii) by procedures outlined in detail elsewhere (Stephens, 1986). Isolation temperatures were 4°C and 25°C, respectively. In all cases, the membranes were removed by extraction at 0°C with 0 ·5% Triton X-100 (or Nonidet P-40) in 30mM-Tris-HCl (pH8), 3mM-MgCl₂, and 0 ·1% 2-mercaptoethanol. The resulting ciliary axonemes were either used directly or else suspended in 30 mM-Tris-HCl (pH 8) and 3mM-MgCl, diluted 1:1 (v/v) with cold glycerol, and stored at —20°C until needed, at which time they were diluted 1:10 with 10 mM-Tris-HCl (pH 8) containing 0 ·1% 2-mercaptoethanol and recovered by centrifugation at 25 000 g for 10 min.

To prepare singlet A-tubules connected together as a set of nine, cilia were dialysed at 0 –4°C against >100 volumes of 1 mM-Tris-HCl (pH 8), 0 ·1 mM-EDTA, and 0 ·1%2-mercaptoethanol for 24-36 h, with at least three changes of solution (Linck, 1973b; Stephens, 1977, 1986). The dialysis-fractionated axonemes were recovered by centrifugation at 45 000g for 15 min; solubilized B-tubule tubulin comprised the bulk of the supernatant protein. To perform the basic thermal fractionation, either intact cilia or A-tubule complexes were suspended in 10 mM-Tris-HCl (pH 8), 1 mM-EDTA, and 04% 2-mercaptoethanol to a concentration of 1 –2 mg ml⁻¹ and heated rapidly to 45°C. Samples were withdrawn at various experimental times and chilled rapidly on ice. After the final sample was taken, all samples were centrifuged at 45 000g for 15 min, the supernatants were withdrawn, and the pellets were washed once with the fractionation buffer and resuspended to the initial sample volume with this same buffer.

For studies of calmodulin-binding stoichiometry, samples of cilia were demembranated with 0 ·5% Triton X-100, 0 ·2M-NaCl, 2 ·5mM-MgClz, 1 mM-dithiothreitol (DTT), and 25 mM-imidazole buffer (pH 7·0), containing 1·0 mM-EGTAand varying ratios of CaClz. The final calcium concentrations were determined with an iterative computer program that corrects for ionic strength, pH, competing divalent cation, and chelation by a nucleotide (J. J. Shoukimas & R. E. Stephens, unpublished). The resulting axonemes were then subjected to a second extraction with the demembranation buffer containing 1 mM-EGTAbut no calcium (described above). The amount of calmodulin in the first and second extract (free and bound, respectively) were determined by SDS-PAGE (below). Axonemes having bound calmodulin were subjected to thermal fractionation, as above, and the time course of calmodulin release was similarly determined.

Samples were analysed on the discontinuous SDS system of Laemmli (1970), using 1·5 mm thick × 10cm long slab gels of uniform 8% acrylamide. Some analyses were performed on gels half this length using a Hoefer Mighty Small II unit. Calmodulin was estimated by quantitative differential densitometry, using its positional shift in parallel samples containing an excess of either 1 mM-CaCl₂ or 1 mM-EGTA, analysed on 5% to 15% acrylamide slabs (cf. Stommel et al. 1982). Staining was with 0 ·025% Coomassie Blue (Serva) by the equilibration method of Fairbanks et al. (1971). For more accurate quantification, some gels were stained by substituting Fast Green FCF at 0 ·1% (Gorovsky et al. 1970). Densitometry was performed with an Ortec 4310 instrument using white light for Coomassie Blue staining and red light for Fast Green. Quantification was by planimetry or, for calmodulin, by direct digitization of data and computer processing. Where gel lane widths varied as a result of sample composition, a simple proportional correction was made.

MAP-free tubulin from the brain of the skate Raja erinacea was prepared by two cycles of in vitro polymerization, using the basic methods developed by Langford (1978) for dogfish brain tubulin. To improve yields, 1 mM-DTT was included in all solutions. Pellets of microtubules were quick-frozen with liquid nitrogen and stored at —80°C until use. In addition, scallop ciliary and sea-urchin (S. droebachiensis) flagellar tubulins were prepared by thermal fractionation using the methods of Linck & Langevin (1981).

Affinity-purified antibodies, prepared against the 47, 51 and 55 (×10³)M_r flagellar tektins (now known as tektins C, B and A, respectively) from the sea-urchin S. purpnratns or against the 51-52 (×10³)M_r tektin from Lytechinus pictus (Linck et al. 1987), were used to determine the presence of antigenically similar ciliary tektins in various fractions. Tubulin was localized using a polyclonal antibody to sea-urchin egg tubulin (Polysciences no. 17870) or a monoclonal antibody to rat brain β-tubulin (ICN no. 63-781). Proteins separated by SDS-PAGE were transferred to nitrocellulose by electroblotting at 60 V for 60 min in a Tris-glycine (pH 8 ·3) buffer containing 0 ·1% SDS and 20% methanol (Burnette, 1981), using a Hoefer TE-22 Mini-Transphor unit cooled to 20°C. Multiple antibodies were tested simultaneously using a Hoefer PR-15,0 Mini Deca-Probe multi-chamber incubation unit. Primary antibodies were visualized by alkaline phosphatase-coupled goat anti-rabbit secondary antibody using the Protoblot system (Promega Biotec).

Thin pellets (<lmm) of samples were fixed at 20°C for 30 –60 min with 2 ·5% glutaraldehyde in 10 mM-Tris-HCl (pH 8) containing 1 mM-EDTA and 0 ·1% 2-mercaptoethanol (and in some cases tannic acid at 0·1 or 1 ·0%), washed with the Tris-EDTA buffer, and post-fixed for 30 min with Karnovsky’s osmium-ferrocyanide (Karnovsky, 1971). The material was stained en bloc overnight with 1% aqueous uranyl acetate, dehydrated with ethanol and propylene oxide, and embedded in Epon-Araldite resin. Thin sections were cut with a Dupont diamond knife and observed and photographed in a Zeiss EM10C electron microscope. To increase contrast further, some sections were post-stained with lead citrate. Dilute preparations were also examined directly by negative staining on carbon-coated Formvar film grids, using 0 ·5% uranyl acetate or 1% uranyl formate.

When axonemes are heated above 40°C, tubulin is rapidly solubilized, leaving most minor protein components as an easily sedimentable material. Temperatures above 50°C tend to denature and precipitate the solubilized tubulin under certain solution conditions. We chose 45°C as a standard temperature for these studies primarily because it permitted a convenient time course for the comparative fractionation of cilia from the three species studied here. Fig. 1 illustrates the time course for solubilization of tubulin from freshly prepared A. irradians ciliary axonemes and from axonemes first dialysed to produce A-singlets. In both cases, maximal tubulin solubilization occurs within 8 min and the same minor proteins remain with the pellet fraction throughout. In the case of fresh axonemes, typically about 80% of the total axonemal tubulin is solubilized at this point. At time points beyond 32min (not shown), more tubulin begins to appear in the pellet fraction, a consequence of denaturation. Essentially identical results can be obtained with cilia stored in glycerol at —20°C for up to 3 years.

In Fig. 2, representative thermal supernatant and pellet fractions are quantified densitometrically, using Fast Green staining of the 8-min sample from fresh axonemes. Even though 79 ·1% of the total tubulin appears in the supernatant fraction, 39 ·3% of the total axonemal protein remains with the pellet. Virtually identical results are obtained from the sum of the dialysis- and thermal-released tubulin versus the thermal-insoluble fraction. In both cases, 21% of the total tubulin remains with the thermal-insoluble pellet, on the basis of the quantification of the α-tubulin chain, for which integration is less complicated by adjacent polypeptides or by the variable splitting of β-tubulin into two equimolar isotypes (Stephens, 1982) that occurs under certain electrophoretic conditions (e.g. the effect is maximal here but minimal in Fig. 7, below). Independently, quantitative immunoblotting with either a polyclonal antibody to sea-urchin egg tubulin or a monoclonal antibody against rat brain β-tubulin gives a 1:4 ratio of tubulin in the remnant fraction versus the thermal supernatant (data not shown).

Molluscan gill cilia are derived from terminally differentiated cells and are known for their stability, while cilia from sea-urchin embryos are constantly resorbed and regenerated. Fig. 3 illustrates the thermal fractionation of ciliary axonemes from 8 droebachiensis, a cold water urchin, and T. gratilla, a warm water species. The latter are relatively more stable while the former are less stable towards temperature than scallop cilia, roughly in accord with their respective environmental temperature. In all three cases, even though the minor axonemal proteins have slightly different (species-dependent) molecular weight distributions, the fractionations approach a final sedimentable product of constant composition, containing most of the minor axonemal protein components and about one fifth of the tubulin.

Electron-microscopic observations reveal that the final product of this thermal fractionation is a remnant retaining much of the symmetry of the original 9+2 axoneme but devoid of microtubules (Fig. 4A,B). In place of the microtubules are nine outer filaments surrounding one or two central filaments. The nine outer filaments are apparently interconnected by nexin linkages and retain disoriented radial spoke material. In longitudinal section, the remnants are fairly uniform, with some evidence of fine axial periodicities. Given the number, location, and orientation of the 9+2 filaments, they must arise from the axonemal microtubules, with the outer nine arising from the region of the A-tubule to which the radial spokes and the centripetal side of the B-tubule are attached. This conclusion is also reached by examination of intermediate stages of melting (Fig. 4C), where the walls of the A-tubules are sometimes reduced to vane-like segments, corresponding in size to about three protofilaments while, after maximum melting, only a relatively more electron-dense filament remains at a radius equivalent to that of the centripetal A-B junction (Fig. 4B). We refer to this most stable region of the A-tubule as the junctional protofilaments. This same basic morphology is observed as a final product, obtained in all three species, whether the fractionation begins with the complete 9+2 axoneme or the dialysis-purified ring of axonemal singlet tubules. In retrospect, many of these same structures are observable, or in the case of nexin linkages, are implicit, in the 9-singlet/l-central pair complex obtained by long-term dialysis (Linck, 1973b), although the preservation was not optimal. We find that ferrocyanide-reduced osmium and en bloc staining with uranyl acetate are critical to the preservation of the remnant structure in the near-absence of tubulin. The relative lability of these structures to normal osmium fixation, but their preservation in reduced osmium, may be related to osmium tetroxidemediated protein cleavage (Maupin-Szamier & Pollard, 1978). ‘

Phase-contrast or dark-field microscopy indicates that these remnants have essentially the same length distribution as the original unfractionated cilia, although repeated centrifugation and resuspension will eventually break them randomly (data not shown). Linck (1973b) showed earlier that cilia fractionated to a cylinder of nine singlet microtubules retained a firmly attached basal plate. This is easily observed in negative stain as a very dense, squarely truncated end (Fig. 5A), but little structure is revealed, due to the thickness of the remnant and the penetration of the negative stain. In regions of the ciliary shaft where the negative stain is fortuitously thin, longitudinal periodicities resulting from radial spoke material are evident (Fig. 5A, inset). A clear break in the shaft was found in less than 1% of our observations. In addition to the basal plate region, the thermally fractionated ciliary remnant must retain at least part of the ciliary tip structure, since the opposite end of the remnant terminates in a tapered region in which the radial material diminishes but the junctional protofilaments continue on into an apparently membrane-associated tip (Fig. 5A). This tip region must correspond, at least in part, to the ciliary cap/distal filament complex described in these same cilia by Dentier (1980). The extension of protofila-merits fully to the ciliary tip is most obvious in sea-urchin ciliary remnants, where the radial material retains the cylindrical structure less tenaciously (Fig. 5B,C). In all three species, the tapered tip region appears to be encased in an amorphous or membranous material. This material is resistant to multiple detergent extractions or to storage of the axonemes in 50% glycerol. Treatment of the remnants with 0 ·l M-MgClz or 0 ·l M-CaC12 does not disrupt the tip region, in contrast to the effect reported for Tetrahymena cilia (Suprenant & Dentier, 1986). This is not totally unexpected, however, considering that the tip structures of protozoan and scallop cilia are rather different (Dentier, 1980). While the tapering tip regions were clearly evident in whole-mount preparations, we were unable to observe, unequivocally, any ciliary tips in embedded material, although basal plate regions were always quite evident in virtually every thin section examined. In all three species studied, the (unfixed) negatively stained material appeared uniformly intact, further suggesting that some disruption must occur as a consequence of fixation and/or embedding.

The ability to fractionate away B-tubule tubulin by dialysis and then melt away the bulk of the A-tubule permits one to localize the tektin proteins within various regions of the outer doublet microtubules. When the final remnant and solubilized tubulin fractions from seaurchin axonemes are assayed by immunoblotting, the remnant retains essentially all of the polypeptides recognized by antibodies to the sea-urchin flagellar tektins A and B and most but not all of the polypeptide recognized by anti-tektin C (Fig. 6A,B,C). Characterization of these ciliary antigens as flagellar tektin homologues is presented elsewhere (Stephens, 1989). The precise identity of tektin antigens in scallop cilia is complicated by the fact that at least four polypeptides can be detected, in various combinations, with sea-urchin tektin antibodies (cf. Linck et al. 1987) but, regardless of nomenclature or homology, essentially all of the tektin A- or B-like antigenicity remains with the remnant. A particularly useful antibody is one against L. pictus 51 –52 (×10³)M_r tektin (cf. Steffen & Linck, 1988), which detects two major and several minor antigens uniquely in the scallop ciliary remnant fraction (Fig. 6D). The two major scallop antigens are also recognized separately by sea-urchin antibodies to tektins A and B. No detectable tektins are found in dialysis-solubilized B-tubulin from scallop cilia (data not shown). The finding that some tektin C is freed during fractionation in all three cases is consistent with earlier observations of its total removal from tektin filaments by urea-Sarkosyl, leaving filaments composed entirely of tektins A and B in 1:1 ratio (cf. Linck & Langevin, 1982; Linck & Stephens, 1987). The finding that tektins A and B are confined entirely to the remnant is consistent with their postulated location within the inner A-B junctional protofilaments (Amos et al. 1986). For reference, the positions of the various tektin antigens are denoted on Figs 1 –3.

Similar ‘melting’ and localization experiments have been performed with the corresponding sperm flagellar axonemes of the two sea-urchin species. A sedimentable remnant fraction, retaining many of the minor components and all of the tektin A- and B-like antigens, is produced in both cases, but only disconnected filaments were evident in the electron microscope (data not shown). This is consistent with the earlier observation that one fundamental difference between rnolluscan or sea-urchin cilia and their corresponding flagella is the manner in which the outer doublets are interconnected (Linck, 1973b; Stephens, 1986).

The constancy of the final product of thermal fractionation, both in terms of SDS-PAGE analysis of minor proteins and in the morphology of the remnant, is rather surprising. Consequently, we chose to investigate three obvious solution parameters (pH, ionic strength, and the presence of divalent cations) for their effect on the final results of the thermal fractionation, illustrated here with scallop cilia.

Selective solubilization of central pair and B-tubule tubulin during dialysis at low temperatures requires a very low ionic strength, the presence of EDTA, and a relatively alkaline pH (Gibbons, 1965a; Linck, 1973b). We found that the thermal fractionation may be performed in a Tris-EDTA medium 10-fold higher in concentration (10mM-Tris, 1 mM-EDTA) than the longterm dialysis fractionation with no measurable difference in results. However, we also found that ionic strengths above 0 ·1 M and pH values below 6·4 prevent tubulin solubilization. Fig. 7A illustrates a 16-min fractionation carried out with either increasing ionic strength or pH. The presence of salt becomes increasingly inhibitory to tubulin solubilization above 25 mM, while increased pH promotes tubulin solubilization. These two factors, however, have little effect on the distribution of most minor proteins that remain associated with the pellet fraction. These proteins remain constant over the ranges studied, the exception being a subset of proteins immediately below tubulin that are proportionately retained along with tubulin.

When the fractionation is performed under standard conditions and the resulting remnant is then subjected to extraction with high salt, the associated proteins are stable to ionic strength in excess of 1 M, using a variety of salts (data not shown). The chaotropic salts NaSCN and KI, however, extract much of the remaining tubulin (cf. Linck, 1976) and disperse or dissolve the remnant but not in any differential manner. Treatment of axonemes with buffers at pH 3 ·5 or below will also solubilize tubulin maximally (cf. Stephens, 1970) but will not produce a remnant of constant composition or of well-ordered morphology (data not shown).

From what is generally known about microtubule stability, it is reasonable to expect that divalent cations such as magnesium and calcium should very effectively stabilize ciliary structure. To investigate this, we performed a 16-min fractionation in the presence of increasing concentrations of these cations, substituted for EDTA in the fractionation medium. Typical results are illustrated in Fig. 7B. Both divalent cations become inhibitory in the millimolar range but calcium ions appear to be about twice as inhibitory as magnesium. The proteins retained by the remnant at less than inhibitory divalent cation concentrations are essentially the same as those found when the fractionation was performed in the presence of EDTA. It is probably not coincidental that essentially the same minimal concentrations of monovalent and divalent cations that stabilize axonemal structure to thermal fractionation also stabilize Tetrahymena axonemes with respect to light-scattering properties and dynein release (cf. Gibbons, 19656). The ability to fractionate cilia thermally in the presence of moderate levels of salt and comparatively high levels of calcium ion should permit various protein binding or release studies to be performed under optimal conditions, for example the binding of ciliary calmodulin.

It was shown earlier that scallop gill cilia contain an unusually high level of calmodulin free within the periaxonemal space, little being bound to the axoneme or found in the cell body proper (Stommel et al. 1982). By carefully buffering the calcium levels during extraction of the membrane plus matrix fraction and by keeping the detergent concentration minimal and the ionic strength relatively high, we now find that more than a quarter of the total ciliary calmodulin remains associated with the axoneme at pCa values in the micromolar range (<1% in 1 mM-EGTA; 4% at 0 ·2 μM-calcium; 13% at 0 ·5, μM; 21% at 1 μM; 29% at 5μM and above). This amount of bound calmodulin is now in excellent agreement with the amount reported in other systems by other workers (Gitelman & Witman, 1980; Otter, 1988). This difference from our earlier work is attributed to the relatively high level and volume of Triton X-100 or Nonidet P-40 that we used, at relatively low ionic strength, to remove the ciliary membrane. It is now known that Triton X-100 can promote the release of bound calmodulin in a calcium-independent manner (Sobue et al. 1981). In addition, the low ionic strength conditions used in our extractions were not optimal; in retrospect, correlating with the observation that 0-2M-salt was required for optimal binding of scallop calmodulin to a fluphenazine affinity column in the presence of Nonidet P-40 (Stommel et al. 1982).

Because Ohnishi and coworkers (1982) report that the primary binding site for Tetrahymena ciliary calmodulin is at or near the interdoublet (nexin) links, we determined the amount of bound calmodulin released during a thermal fractionation carried out in the presence of 0-1 mM-calcium, a value that should promote calmodulin binding yet not inhibit thermal fractionation or precipitate tubulin (cf. Fig. 7B). Starting with axonemes prepared by membrane extraction with 0 ·1 mM-calcium present, we find that most detectable calmodulin (>75%) is released concomitant with and proportional to the solubilization of tubulin. We have also performed the initial thermal fractionation in the micromolar calcium range and then increased calcium to millimolar levels before centrifugation to see if freed calmodulin would rebind to the remnant. The results were not significantly different from those done with relatively high calcium throughout. Since moderate ionic strength inhibits the thermal fractionation, salt was not included as a variable in such calmodulin release experiments. Bound calmodulin remains bound to the axoneme when it is transferred to a lower ionic strength medium after detergent demembranation and, hence, the continued presence of a moderate salt concentration was not considered critical.

To test the ability of the ciliary remnant to initiate microtubule polymerization, we polymerized MAP-free skate brain tubulin in vitro over a wide range of tubulin concentrations, using the scallop ciliary remnant as a ‘seed’. Whether substantially below, at, or well above the critical concentration for microtubule formation, the ciliary remnant does not promote microtubule reassembly within the remnant (Fig. 8). This is in surprising contrast to fragments of axonemes, which will easily promote singlet microtubule formation from the free ends of A-tubules (e.g. see Binder et al. 1975). In our case, neither doublet nor singlet microtubules are reconstructed in the remnant proper but abundant free singlet microtubules do form in the surrounding medium, at or above the critical concentration. The remnants consistently become considerably more electron-dense in the presence of repolymerizing tubulin, with the junctional protofilament region becoming more defined, but well-ordered tubular structures are not detectable. Although some occasional A-singlets can be detected within the remnant structure, their number is not significantly different from those (basal) microtubules resistant to the fractionation process. In addition to skate brain tubulin, similar experiments were performed with thermally depolymerized axonemal tubulin from both scallop cilia and sea-urchin sperm flagella (cf. Linck & Langevin, 1981). In all cases, while the density of the remnant was increased substantially, no obvious axonemal microtubules were reconstituted.

The results presented here indicate that the ciliary axoneme can be fractionated into a ‘skeleton’ of its original structure. This skeletal remnant retains a 9+2 or 9+1 arrangement of filaments in place of the microtubules, and the outer nine filaments are interconnected by presumed nexin linkages and radial spoke components. Biochemically, the remnant retains about 21% of the total tubulin and it is composed of most of the minor axonemal structural proteins, perhaps most significantly of the three ciliary tektins.

The precise location of these tektins within the ciliary structure has not been determined but, by reference to the localization of tektins in the protofilament ribbon domain in flagellar A-tubules (Linck & Langevin, 1982; Linck et al. 1985), the nine outer remnant filaments must be derived from the A-tubule and must be situated at or near the inner A-B junction and near the point(s) where radial spokes and nexin bridges associate with the doublet (cf. Amos et al. 1986). The general location of these tektins is implied by observations during the time course of melting, by the final remnant morphology, and by the near-quantitative immunochemical localization of the tektins in the remnant fraction. It is far less likely that tektins are found in the major portion of the A-tubule wall that serves as a partition with respect to the B-tubule, since the bulk of this domain is absent in the final remnant. Tannic acid staining has not shed light on this question, since no obvious protofilaments are enhanced in the remnant, probably as a consequence of the high degree of disorder along the protofilament axis after thermal fractionation. Some structures also remain near the centre of the axonemal remnant, possibly corresponding to tektin-associated stable protofilaments derived from one or both of the central pair members (cf. Steffen & Linck, 1988), potentially complicating protein quantification and localization arguments.

The exact stoichiometry of tubulin to tektin in the outer doublets is also uncertain but some interesting parallels can be drawn with earlier work. Our data show that about 21% of the total ciliary tubulin remains associated with the remnant. This theoretically corresponds to 5 ·4 protofilaments remaining at each outer doublet locus and also to 23 ·6% of the outer doublet tubulin. An independent calculation, based on the densitometrically measured amount of cr-tubulin solubilized by dialysis versus that subsequently solubilized by thermal fractionation (an experimental ratio of 1 ·30 to 1 ·00) is consistent with the solubilization of 7 ·9 protofilaments from each A-tubule by thermal fractionation, leaving 5 ·1 from each with the remnant. This calculation assumes solubilization of nine B-tubules of 10 protofilaments each plus one central pair member of 13 protofilaments in the dialysis fraction and the other central pair member plus some unknown number of protofilaments from the nine A-tubules, containing 13 protofilaments each, in the thermal fractionation. Earlier bulk fractionation procedures showed that 24% of the outer doublet tubulin remained associated with tektin ribbons derived from seaurchin sperm flagella (Linck, 1976). These ribbons appeared morphologically to consist of three protofilaments (which could account for only 13% of the tubulin), prompting the suggestion that each outer doublet might contain two tektin-tubulin protofilament ribbons (i.e. six tubulin-containing protofilaments and hence 26% of the outer doublet tubulin). This value is not significantly different from the value determined here for scallop ciliary remnants, particularly considering the possible cumulative experimental errors in quantification and the fact that complex rnolluscan gill ciliary remnants are being compared with simple sea-urchin sperm flagellar ribbons.

The isolation of a stable structure, devoid of microtubules, retaining both the ninefold symmetry of the ciliary axoneme and most of the minor protein components, would suggest that the cilium is built linearly in a very integrated fashion, in essence by the assembly of a framework or skeleton upon which, or into which, at least the nine outer doublets must assemble. An interesting parallel situation exists in the case of the ‘centriolar rim’, a structure that maintains the configuration of basal bodies in the absence of microtubules (Fais et al. 1986). Furthermore, Gavin (1984) has reported the in vitro formation of ninefold symmetric structures from extracts of Tetrahymena basal bodies, some of which contain singlet microtubules. The finding that we have not been able to assemble singlet microtubules onto the ciliary remnants suggests, but certainly does not prove, that cilia may form by a coordinated co-assembly of both major and minor protein components. This basic concept was proposed for ciliogenesis in sea-urchin embryos, wherein one protein, now known to be tektin A, is made quantally and utilized fully, such that its requisite co-assembly into the outer doublets would consequently limit ciliary length (Stephens, 1977). Correlation of ciliary length and elongation with the quantal synthesis of tektin A lends support to this idea (Stephens, 1989).

The fact that the remnant is continuous throughout the length of the cilium, plus the fact that it extends all the way to the ciliary tip, is consistent with the immunolocalization of the three tektins uniformly along all nine outer doublet microtubules, including the A-singlets that extend to the tip, as observed in sea-urchin flagella (Steffen & Linck, 1988). It has been reported that the ciliary tip structures remain firmly attached to growing outer doublets during ciliary regeneration in Tetrahymena (Dentier, 1980). Assuming that the same situation exists in scallop and sea-urchin cilia, this would imply that the tektin filaments may be intimately involved in the assembly process, possibly functioning within the assembly site itself. The relationship, if any, between the tektin-containing filaments that extend to the ciliary tip (Steffen & Linck, 1988) and the distal filaments observed as part of the A-tubule luminal plugs that evidently mediate membrane interaction (Dentier, 1980) is at present unclear..

One of our original purposes in carrying out these fractionation studies was to determine the location of ciliary calmodulin, a protein that occurs in unusually high amount in scallop cilia (Stommel et al. 1982) and which is purported to bind in the vicinity of interdoublet linkages (Ohnishi et al. 1982). We find no evidence that the bulk of the axoneme-bound calmodulin is binding to the ciliary remnant, suggesting that nexin linkages and the other minor architectural proteins are not the primary binding sites for calmodulin. Although the release of calmodulin parallels the release of tubulin, this is not proof of calmodulin binding to tubulin per se, since an intermediate binding protein could be involved. The mole ratio of bound calmodulin to tubulin dimer in the whole axoneme is about 1:60, a value that does not correlate with any obvious microtubule or axonemal structural features, although this same ratio is found consistently in other cilia and flagella as well (cf. Otter, 1988). We cannot rigorously exclude the possibility that some calmodulin may bind to nexin bridges, since one calmodulin per interdoublet linkage would account for only one calmodulin per nearly 280 tubulin dimers and at least this amount remains with the remnant after fractionation (but so also does a proportionate amount of tubulin).

The precise location of the tektins within or amongst the tubulin protofilaments and the interaction of these unique protofilaments with the various architectural proteins of the axoneme must remain questions for future work. It should be clear, however, that these complex interactions must be important determinants of ciliary structure, given their persistence and stability after removal of most of the structure of the 9+2 microtubules under a wide variety of ionic conditions.

This work was supported by USPHS grants GM20644 (to R.E.S.) and GM35648 (to R.W.L.) from the National Institute of General Medical Sciences. We thank Drs Jonathan J. Shoukimas and D. Wesley Corson for their assistance in the quantification of the calcium-calmodulin data, Dr Robert E. Kane for his hospitality and assistance at the University of Hawaii’s Kewalo Marine Laboratory, and Dr Lionel I. Rebhun for pointing out to us that the thermal fractionation of tropical sea-urchin axonemes will yield a ninefold remnant structure, thus initiating some of our early experiments.