Studies Of Tektin Filaments From Flagellar Microtubules By Immunoelectron Microscopy

It has been shown by various workers that sea-urchin sperm flagellar doublet microtubules can be dissociated in a number of ways to produce chemically resistant ribbons of two to four protofilaments (Linck, 1976; Meza et al. 1972; Witman et al. 1972). These are thought to form one or more parts of the A-tubules of the outer doublets, in particular the region where the C-shaped B-tubule makes a junction with the A-tubule wall (Linck, 1982). The ribbons have been further fractionated with chemical agents to give highly insoluble ‘tektin’ filaments (Linck & Langevin, 1982).

In each species studied, the tektins consist of several distinct polypeptides varying somewhat in molecular weight (between 45 and 70 (×10³M_a) and isoelectric properties. Their general properties suggest that they are much more closely related to the proteins of intermediate filaments (see, e.g., Fuchs & Hanukoglu, 1983) than to tubulin: they are highly insoluble in solutions containing Sarkosyl and urea; their sensitivities to limited proteolytic cleavage by Staphylococcus aureus protease are different from that of tubulin (Linck & Langevin, 1982); as a group they appear as fine fibrils with a high (71%) α-helical content, as shown by circular dichroism (Linck & Langevin, 1982); finally, X-ray fibre diffraction of the purified filaments gives patterns strongly characteristic of coiled-coil α-helical proteins (Beese, 1984).

Relatively stable components have been detected also in flagellar central singlet microtubules (Linck et al. 1981) and in microtubules from mitotic spindles (Jensen & Bajer, 1973). The latter may consist of the 55× 10³M_?. non-tubulin protein, found by Hays & Salmon (1983) to be the major component of the calcium-resistant residue of isolated mitotic spindles from Strongylocentrotus purpuratus eggs. There is a possibility, therefore, that tektin-like polypeptides are a general feature of microtubules.

The tektins of 5. purpuratus flagellar microtubules consist of a group of three or more polypeptides that run on sodium dodecyl sulphate (SDS)–polyacrylamide gels in much the same region as α- and β-tubulin. Fig. 1, lanes a-c, shows the polypeptide compositions of the remaining polymers when axonemes are chemically disassembled. When A-tubules (Fig. 1, lane a) are treated with 0·5% Sarkosyl, ribbons of two to four protofilaments (Fig. 1, lane b) remain (Linck, 1976) but only a fine filamentous fraction (Fig. 1, lane c) is insoluble in 0·5% Sarkosyl plus 2?-urea (Linck & Langevin, 1982). The tektin bands can be seen in each case. After a second extraction in Sarkosyl—urea, the 47, 51 and 55 (× 10³M_?) bands together constitute more than 95% of the Coomassie-stained protein. The ribbons of protofilaments contain other proteins in addition to tubulin and the three main tektin bands, in particular a pair of bands having apparent molecular weights of 77 and 83 (×10³) (Linck & Langevin, 1982); their properties have not as yet been investigated. Other species give rise to different sets of polypeptide bands on SDS–poly- acrylamide gels but the filaments all look similar in the electron microscope (Linck & Langevin, 1982).

Mouse antibodies that were raised against the Sarkosyl-urea-purified tektin filament complex were affinity purified using the antigen coupled to Sepharose 4B. On nitrocellulose blots of SDS–polyacrylamide gels, the purified anti-tektin antibodies reacted with all three main tektin bands from S. purpuratus, but disproportionately (Fig. 1, lanes e, h). They also cross-reacted with similar but not identical sets of proteins in SDS-polyacrylamide gel blots of axonemes from cilia and flagella of other sea-urchin species, but with reduced affinity (Linck et al. 1985).

Anti-tubulin antibodies did not bind to the tektin bands but did reveal a residual amount of tubulin in the original preparation (Fig. 1, lane i), below the level of detection with Coomassie Blue staining. Purified anti-tektins did not bind to tubulin on nitrocellulose blots (Fig. 1, lane k), but to be quite sure that the immunoglobulin G (IgG) fraction used for the structural work contained no trace of anti-tubulin activity, it was passed through a tubulin—Sepharose column.

To investigate the relationship between tektins and intermediate filaments, we also applied the ‘universal’ anti-intermediate filament (IF) monoclonal antibody (Pruss et al. 1981) to gel blots. No reaction was detectable when the antibody was applied to blots from whole axonemes. However, on blots of purified filament protein, the 55 and 51 (× 10³)M_?. tektin bands weakly but reproducibly bound the antibody, while the 47 ×10³M_?band sometimes showed weak binding (Fig. 1, lane n). The antibody also revealed traces of some more slowly running material whose presence was not evident from Coomassie Blue staining. Though weak, the anti-IF antibody binding to both the tektin bands and the other material seems to be specific, since under the same conditions other antibodies such as anti-tubulins did not bind. This result suggests that tektins may contain a stretch of sequence similar but not identical to the common intermediate filament epitope.

Our purified anti-tektin antibodies did not appear to recognize mammalian intermediate filament proteins. This is not surprising in view of the species-specificity of the sera (see below). However, they do not even bind to nitrocellulose blots of mitotic spindles isolated from fertilized eggs of the same species (Amos et al. 1985), even though Hays & Salmon (1983) have shown that such spindle preparations include a 55×10³M_r non-tubulin protein that may belong to the intermediate filament class. This lack of cross-reactivity is borne out by the results obtained by indirect immunofluorescence microscopy.

Sea-urchin sperm fixed with formaldehyde in sea water were stained all along their lengths after incubation with purified mouse anti-tektins, followed by FITC antimouse IgG (Linck et al. 1985; Amos et al. 1985). Similar results could be obtained for demembranated, purified axonemes (Fig. 2A) provided they were fixed in a suitable manner. Fixation simply in methanol cooled on solid CO₂ worked well. Fixation by formaldehyde in phosphate-buffered saline (PBS), followed by cold methanol, gave reasonable results. However, if the axonemes were kept during the early stages of aldehyde fixation in the axoneme-washing buffer of Linck & Langevin (1981), which includes 5 mM-MgSO₄ and 0·5 m?-EDTA, and is known to give good structural preservation at the electron microscope (EM) level, they did not stain with FITC. Anti-tubulin antibodies produced positive results for axonemes fixed in any of these ways, whereas staining did not occur under any conditions if pre-immune sera from the tektin-injected mice were used. These results suggested to us that a denaturation or unmasking of tektins is necessary for anti-tektin binding.

Staining could be obtained for axonemes from all species of sea urchin that were tried. However, the minimum concentration of anti-tektin required for a positive result was a hundred times higher for Lytechinus pictus. Echinus esculentus and Psammechinus miliaris than for S. purpuratus, the source of the original antigen.

The purified antibodies were also used to investigate a variety of structures containing microtubules. Since the antibodies showed reduced reactivity even when applied to axonemes of sea-urchin species other than 5. purpuratus, we concentrated on material from echinoids, such as mitotic spindles and cytasters from eggs, and cilia and cytoplasmic microtubules of embryos (Amos et al. 1985).

Ciliated embryos were obtained by fertilizing L. pictus eggs and leaving them to develop in artificial sea water. They were fixed in cold methanol, following the method that gave good anti-tektin staining for axonemes. After anti-tubulin treatment, the embryos were brightly fluorescent. Stained cilia were detected on all surfaces of the embryo and in its interior; each ectodermal cell contained a brightly stained star-shaped structure (Fig. 2B), presumably an array of cytoplasmic microtubules radiating out from a centrosome. Anti-tektin, on the other hand, stained only the cilia (Fig. 2C). This result shows that cilia, whose proteins differ from those of flagella in many cases (Linck, 1973), have components very like flagellar tektins. The cytoplasmic microtubules apparently do not.

This conclusion was supported by results of immunofluorescence experiments with spindles isolated from both S. purpuratus and L. pictus fertilized eggs or with multiple asters induced in unfertilized eggs (see Amos et al. 1985). No staining was observed with anti-tektin antibodies, no matter how the specimens were fixed, although anti-tubulin gave positive results in all cases (e.g. Fig. 2D).

To study the molecular organization of tektin filaments in microtubules by electron microscopy, we used immunolabelling with 5 nm gold particles and uranyl acetate as negative stain (Linck et al. 1985). Gold-labelled anti-tubulin gave quite satisfactory results, though the gold appeared preferentially on singlet A-tubules and on protofilament ribbons (Fig. 3F,G); the tubulin in doublet microtubules may perhaps be masked by accessory proteins.

When preparations of S. purpuratus doublet or singlet microtubules, or protofilament ribbons, suspended in the axoneme buffer referred to above, were applied to carbon-coated grids and incubated with antibodies, first with the affinity-purified mouse anti-tektins and then with gold-labelled rabbit anti-mouse IgG, none of the specimens showed any sign of gold decoration. To ‘unmask’ the tektins, we briefly extracted microtubules on the grid with Sarkosyl–urea, before incubating with antibodies. The extraction in situ left behind filaments 2–3 nm in diameter, often twisted together into loose bundles (similar to the specimens shown in Fig. 4). It is not clear whether all the filaments in a bundle came from the same or different microtubules. After antibody treatment, the filaments were heavily decorated with gold. In control experiments, where no first antibody was used, the gold-labelled second antibody did not bind to the filaments.

In another series of experiments, P. miliaris axonemes were resuspended in a simple buffered saline instead of the axoneme buffer. Then the microtubules and protofilament ribbons all broke down into filaments and amorphous aggregates during the antibody incubations; the filaments became labelled with gold as in Sarkosyl—urea-extracted preparations. However, when 20 µM-taxol was included in the saline, the doublet microtubules disassembled only as far as protofilament ribbons during the incubations. Gold labelling was then mainly restricted to the broken ends of the ribbons, from which thin filaments 2–3 nm in diameter were often seen protruding (see Fig. 3A–E). These observations strongly support the theory that tektins are masked by tubulin in the intact microtubules.

A fusion of myeloma cells with spleen cells from one of the mice immunized with tektins produced several positive wells as assayed by immunofluorescence microscopy of sea-urchin axonemes. None of the supernatants stained mitotic spindles or cytoplasmic microtubules in HeLa cells. The supernatant from the cells in one well, which gave much brighter staining than the others, was used for immunoelectron microscopy of tektin filaments prepared by Sarkosyl–urea extraction of P. miliaris axonemes. The filaments became labelled with gold as before, except that now the labelling frequently appeared to be periodic (Fig. 4). Because the gold particles were attached to a secondary antibody, which amplifies any binding, the locations of the attachment sites of the primary antibody are less certain than if a gold-labelled primary antibody had been used. However, the small quantities of antibody available made the latter option impossible.

An estimate of the spacing of the binding sites was obtained by measuring the distances between the centres of adjacent clumps of gold particles. Fig. 5 shows the results for 54 intervals each defined by discrete clumps of gold. The uncertainty in each measurement is the sum of the diameters of the two clumps enclosing the gap. The distribution has a peak at around 48 nm. Since the highest values probably represent multiples of the basic repeat distance, they were not included in calculating the average. The actual distance between equivalent antigenic sites could be smaller than 48 nm; for example, one-half or one-third of this value. The clumps of label are sufficiently large for spacings of this order to be obscured. However, a periodicity of around 48 nm seems to be the most likely.

The evidence suggests that most of the tektin polypeptides are assembled as continuous 2–3 nm filaments, though there may be fine projections not detected by electron microscopy. A length of about 48 nm per polypeptide in the form of a coiled α-helix would correspond to around 340 amino acid residues (roughly 37 × 10³M_r); this value agrees rather well with the estimated 71% α-helix for tektins. There is sometimes some indication in electron-microscopic images (e.g. Fig. 4A) of globular domains along the filaments, with a periodicity of around 3 nm, significantly less than the 4nm tubulin monomer repeat. It is not clear whether these are due to components other than the main tektin polypeptides, to the non-α-helical regions of the latter or simply to variations in charge distribution along a coiled coil, being shown up by the binding of the uranyl acetate stain.

We have not yet determined where exactly the tektin filaments are located in the A-tubule wall. It is clear that the antigenic sites are blocked in the intact structure, either by tubulin or some of the other associated proteins, such as the 77 and 83 (× 10³)M_r polypeptides of S. purpuratus. It seems unlikely that the antigenic sites were exposed in our experiments by a drastic conformational change of the tektin filament itself, as the conditions used to produce staining were mostly quite mild. Thus, the most probable arrangement is a tektin filament running along a groove between two protofilaments, as in the model drawn in Fig. 6. Alternatively, tektins might themselves form certain protofilaments, in association with other non-tubulin polypeptides. Either way, they (and any side projections) probably have a stabilizing effect on the region of the wall into which they are incorporated. We do not know whether there is more than one stable protofilament ribbon per A-tubule or whether the different tektin polypeptides form homo- or heteropolymers. Answers to such questions should come from more detailed structural studies with a wider range of specific antibodies.

For a filament running straight alongside a tubulin protofilament, a periodicity of approximately 48 nm would correspond fairly closely with the axial extent of 12 tubulin monomers. This would be half of the overall axial repeat of doublet microtubules as defined by the groups of radial spokes and other accessory structures (see Fig. 6). Indeed, the tektins, being longitudinal elements, are quite likely to be involved in ordering the axial spacing of such components, as suggested by Linck (1982). The 29% ‘random chain’ portions of the polypeptides may perhaps extend sideways to organize laterally distant accessory structures.

A further aspect of the possible role of tektins as longitudinal rulers along microtubules is that they may also define the ultimate length of the axonemes (Amos et al. 1985). Stephens (1977) made an extensive study of protein synthesis during cilio- genesis in sea-urchin embryos and found that a small number of minor components are synthesized in a short burst, either after fertilization or after experimental deciliation, whereas tubulins and dyneins are not newly synthesized for the purpose. The newly synthesized polypeptides were incorporated almost totally into the growing cilia. It is interesting that the major species was component 20, which is found in the Sarkosyl-resistant residue of the A-tubule and therefore is probably a component of the ciliary tektins. As suggested by Stephens, the apparently limited supply of this component might be the controlling factor in limiting the growth of the cilia.

It seems likely that microtubules in other cellular structures also possess some means of controlling their lengths in the presence of pools of tubulin required for different purposes. Although the negative results for spindles and cytasters could mean that tektins are absent from the constituent microtubules, there may be tektin- like proteins present but with no strong antigenic determinants in common with flagellar tektins. Again, a wider range of antibodies may answer this question. Since other microtubules are required to be much less stable than flagellar microtubules, it would not be surprising if they had associated proteins similar to tektins in some respects but different in others. The calcium-insoluble filaments described by Hays & Salmon (1983) may represent such components.

Finally, the question arises as to how closely related tektins are to bona fide intermediate filament proteins. Work in progress on the amino acid sequences of tektins should finally resolve this question. As emphasized before, many of their properties are similar and the weak reaction with the anti-intermediate-filament antibody suggests some homology in the sequence. We have discussed the necessity for tektin filaments to match up with the axial periodicities of the tubulin protofilaments. The nominal 4nm monomer periodicity in microtubules has actually been measured as 4·05–4· 10 nm (Mandelkow et al. 1977; Baker & Amos, 1978), probably the most accurate estimate, 4·09nm, being that of Wais-Steider et al. (1986). McLachlan & Stewart (1982) detected a regular pattern of charged residues in the primary sequences of the intermediate filament proteins desmin and vimentin with a period of 28 residues and pointed out that the axial extent of this repeat is similar to the tubulin monomer repeat in a microtubule. The match is actually even better than they thought. In a coiled coil, a 28-residue periodicity would correspond to an axial distance of 4·O8nm, a value remarkably close to the tubulin monomer repeat (see Fig. 7). Also, the total length of an individual intermediate filament protein monomer is thought to be around 44 nm (Crewther et al. 1983), only slightly shorter than our estimate of 48 nm for the length of a tektin monomer. Desmin and vimentin may not themselves interact directly with tubulin, but they may possibly be derived from an ancestral protein that evolved within a microtubule wall.

We are grateful to Dr B. H. Anderton for a generous supply of anti-intermediate-filament antibody; also to Dr J. V. Kilmartin for advice and gold-labelled anti-tubulin.