Tektin B1 from ciliary microtubules: primary structure as deduced from the cDNA sequence and comparison with tektin A1

Tektins were originally identified as a set of proteins com-prising relatively insoluble, 2-3 nm diameter filaments that remain after the extraction of sea urchin sperm flagellar microtubules with chaotropic solvents (Linck and Langevin, 1982). Three tektins have so far been characterized: tektin A (∼53 kDa), B (∼51 kDa) and C (∼47 kDa). Tryptic pep-tide mapping (Linck and Stephens, 1987) and immunolog-ical studies (Linck et al., 1985; Steffen and Linck, 1988) demonstrated that the three tektins were different but related polypeptides and that they were unrelated to tubulin. Initial biochemical and structural data suggested that these novel proteins were similar to intermediate filament protein (IFP) from vertebrates in terms of their relative insolubility, mol-ecular masses, and fibrous, α-helical substructure (see Linck, 1990). Immunological studies further substantiated the similarity between tektins and IFP (Chang and Piperno, 1987; Steffen and Linck, 1989a). It was surprising, there-fore, that the first sequence reported for tektin A (Norran-der et al., 1992) revealed only a weak homology with IFP. The organization of tektins in microtubules is not known in detail, but tektins probably interact directly with tubulin. Immunofluorescence microscopy has demonstrated that tek-tins A, B and C are present in all nine flagellar doublet microtubules and possibly in the central pair microtubules (Steffen and Linck, 1988). Immunoelectron microscopy suggests that tektins form extended polymers in the walls of ciliary and flagellar microtubules (Linck et al., 1985). Sarkosyl detergent extraction of cilia and flagella yields insoluble ribbons of approximately three protofilaments, composed of tubulin and tektins in a close molar ratio, suggesting a direct molecular interaction (Linck, 1976). Finally, at least one form of the insoluble protofilament rib-bons is localized to that part of the A-tubule to which the B-tubule and nexin links join (Linck, 1976; Stephens et al., 1989).

Immunological evidence suggests that tektins or tektin-related polypeptides are present in other microtubule sys-tems. Besides staining flagellar axonemes, tektin antibodies also stain basal bodies in sea urchin sperm (Steffen and Linck, 1988). This result is not unexpected, since the dou-blet tubules assemble from the plus ends of the basal body triplet microtubules (Alberts et al., 1989; Allen and Borisy, 1974), and therefore, some proteins might be common to both doublet and triplet microtubules. Perhaps more sur-prisingly, tektin antibodies also stain centrioles in a variety of mammalian cells, including HeLa cells (Steffen and Linck, 1988), suggesting that tektins could be evolutionar-ily conserved components of centrioles. Tektin-like com-ponents have also been detected in centrosomes and in spin-dles and midbodies of dividing cells (Steffen and Linck, 1989b, 1990, 1992, and unpublished observations).

We report here the cloning and sequencing of a cDNA for tektin B from sea urchin embryos. There were several reasons for this approach. First, sea urchin blastula embry-onic cilia had been shown to contain tektins both by immunofluorescence microscopy and by SDS-PAGE immunoblotting, using antibodies raised against sperm fla-gellar tektins (Amos et al., 1986; Linck et al., 1987). Second, Stephens (1977) had shown that a ciliary micro-tubule protein of ∼55 kDa is synthesized de novo at the onset of ciliogenesis in sea urchin blastulae; this 55 kDa polypeptide has since been identified as tektin A (Stephens, 1989) by crossreaction with a polyclonal antibody against flagellar tektin A (Linck et al., 1987). On the basis of these findings and having had success with cloning tektin A cDNA in this manner, we chose to screen a λZapII cDNA library from Strongylocentrotus purpuratus sea urchin blastula with antibodies against flagellar tektin B.

A λZapII cDNA library was constructed commercially (Strata-gene, Inc., La Jolla, CA) from mRNA of Strongylocentrotus purpuratus sea urchin blastula embryos, as prepared by Norrander et al. (1992). The cDNA library was screened with polyclonal anti-bodies specific to S. purpuratus sperm flagellar tektin B (Linck et al., 1987), according to established protocols (Huynh et al., 1985; Young and Davis, 1983). Positive plaques were identified by incu-bation of filters in a solution of alkaline phosphatase-conjugated goat anti-rabbit IgG (Stratagene, Inc.) for 1 hour, followed by color development with nitro blue tetrazolin (NBT) and 5-bromo-4-chloro-3-indolyl phosphate (BCIP) provided in the commercial kit.

Procedures for the purification of phage and plasmid DNA and the procedures for RNA blots are based on established techniques with some modification (Sambrook et al., 1989). pBluescript plas-mids containing the cDNA inserts were excised from λZapII clones using the helper phage R408. The probe for northern blots was prepared by digesting the purified tek-B1 cDNA clone with EcoRI, isolating the tek-B1 insert and then labeling by nick-trans-lation with [³²P]dCTP, as described in the manufacturer’s proto-col (BRL, Bethesda, MD). Hybridizations were carried out at 42°C for 18 hours in a solution containing 50% deionized formamide, 5× SSC (1 × =150 mM sodium chloride, 15 mM sodium acetate), 5× Denhardt’s (1× =0.02% Ficoll, 0.02% polyvinylpyrrolidone, 0.02% bovine serum albumin), 20 mM NaHPO₄, pH 6.5, 0.1% SDS, and 50 μg/ml sheared, denatured salmon sperm DNA and 10⁶ cpm/ml denatured radiolabeled cDNA probe. Filters were washed four times 10 minutes in 2× SSC, 0.2% SDS at room temperature, followed by two 15-minute washes in 0.1× SSC, 0.1% SDS at 50°C.

The nucleotide sequences of the clones were determined using the dideoxy chain termination method of Sanger et al. (1977). Single-stranded DNA templates were prepared by infecting cells harbor-ing the plasmid pBluescript SK containing cDNA inserts with VCS-M13 filamentous bacteriophage, and then by isolating the resultant progeny phagemids (Sambrook et al., 1989). Partial clones were prepared by constructing overlapping deletion sub-clones based on controlled digestion of the double stranded DNA with exonuclease III (Putney et al., 1981), as described by Henikoff (1984, 1987) with modification. The United States Bio-chemical Corporation (Cleveland, OH) Sequenase kit (version 2.0) and the Promega (Madison, WI) TaqTrack sequencing system were used to prepare the chain termination reactions with [³⁵S]dATP, as described in manufacturer’s protocols. We sequenced either single-stranded DNA or double-stranded DNA from cDNA subclones in plasmid pBluescript SK (Stratagene) containing the full-length cDNA (tek-B1) or partial clones (tek-B2, tek-B3).

Purification of inclusion bodies of bacterially expressed protein was performed according to the procedure of Marston (1987; Marston et al., 1984) and Rüther and Müller-Hill (1983) with mod-ification. Tek-B1 cDNA/pBluescript SK⁻ (Stratagene, Inc.) in LB media with a final concentration of 150 μg/ml ampicillin and 12.5 μg/ml of tetracycline was incubated at 37°C for 4 hours. Expression protein was induced with 1 mM IPTG and incubation overnight at 37°C with shaking. 500 ml of cell culture was cen-trifuged at 500 g for 15 minutes at 4°C. Each gram of cells was suspended in 3 ml lysis buffer (50 mM Tris-HCl, pH 8.0, 1 mM EDTA, 100 mM NaCl), 4 μl of 50 mM PMSF, 80 μl lysozyme (10 mg/ml), and 4 mg deoxycholic acid, and sonicated three times 10 seconds. Escherichia coli DNA was digested by adding 20 μl DNase I (1 mg/ml) and incubating at 37°C for 30 minutes. Cell lysates containing inclusion bodies were pelleted by centrifugation at 12,000 g for 15 minutes at 4°C. Inclusion bodies were fur-ther treated by resuspension in 15 ml lysis buffer with 0.5% Triton X-100, 10 mM EDTA, pH 8.0, and pelleting as above. The result-ing protein was analyzed by SDS-PAGE immunoblotting (Towbin et al., 1979), using affinity-purified, polyclonal anti-(S.p.)-tektin B (Linck et al., 1987) and mouse monoclonal (#52-3-3) anti-tektin B (W. Steffen and R. Linck, unpublished observations).

The pBluescript SK phagemid (Stratagene, Inc.) contains 111 nucleotides coding for β-galactosidase (37 amino acid residues, accounting for an increase of 3,863 kDa in the fusion protein), followed by the EcoRI site into which our cDNA is cloned. By definition, our screening procedure isolated a clone with the correct reading frame to produce a fusion protein recognized by affinity-purified, polyclonal antibody to S. purpuratus tektin B. Thus, any 5′ untranslated region in the cDNA sequence without any stop codons would be readable and would also increase the apparent molecular mass of the fusion protein.

Sperm flagellar tektins were purified by N-lauryl-sarcosinate/urea extraction and reversed-phase (RP) HPLC, according to Linck and Stephens (1987). The purity of tektin B was assessed by SDS/PAGE. Tektin B was dissolved to ∼2.5 mg/ml in 0.4 ml 70% trifluoroacetic acid and treated with cyanogen bromide, at ∼30 μmoles of CNBr/mg protein. The mixture was flushed with N₂ and incubated at room temperature for 24 hours in darkness. After termination of the reaction by addition of excess water, the sample was evaporated to dryness. Samples were dissolved in 0.4 ml 6 M guanidine, 25 mM Tris, pH 8.5, 1 mM EDTA, 5 mM DTT, incubated for 30 minutes at room temperature, and subjected to RP-HPLC for peptide separation. Fractions containing pure pep-tides were collected and sequenced by automated Edman degradation on an Applied Biosystems 470 peptide sequencer, in the Microchemical Facility, University of Minnesota.

The sequence was analyzed using the program ANALYSEP (Staden, 1988), based on the studies of Robson and Suzuki (1976), in order to predict stretches of sequence that are more likely to form α-helix than any alternative secondary structure. The potential for α-helical coiled-coil formation was determined by the method of Lupas et al. (1991), which scores any stretch of α-helix for the presence of hydrophobic residues in the first (a) and fourth (d) positions of a heptapeptide repeat. Probability of coiled-coil formation was calculated for a 28-residue window.

Different stretches of sequence were compared using the pro-gram DIAGON (Staden, 1982) with a proportional matching cal-culation, first described by McLachlan (1982). The method of scoring depends on a table of amino acid similarity based on the substitution rates in known structures of families of related pro-teins (Dayhoff, 1972). The program produces a two-dimensional plot in which the x axis represents one sequence (tektin X) and the y axis represents the same or another sequence (tektin Y). Every point (x, y) is assigned a score that corresponds to the level of similarity between sequence characters over a chosen SPAN centered on point x in sequence X and point y in sequence Y, allowing no insertions or deletions. For analysis of tektin sequences, we chose two different span lengths: 11 residues, the default value for short-range sequence comparison, and 49 residues for longer-range comparison. For each point (x, y), the program calculates the probability of obtaining the same final score by chance with two infinitely long sequences of the same composition as those being compared. We set different minimum values for plotting the probabilities for different spans, in order to show the peaks clearly in each case; matches shown over spans of 11 residues have a probability of <0.0006 of occurring by chance, and matches shown over spans of 49 residues have a prob-ability of <0.00001.

We obtained several positive cDNA isolates by probing a λZapII cDNA library from Strongylocentrotus purpuratus. Of these, clone tek-B1 has a size of 2,621 bp, sufficiently large to be a near-full-length cDNA. Two other clones were also isolated, tek-B2 and tek-B3, with sizes of ∼1,300 bp and ∼950 bp, respectively. Radiolabeled tek-B1, used to probe poly(A)⁺ RNA isolated from sea urchin blastula, hybridizes with a single mRNA band of ∼2,700 bases (Fig. 1). Since the pBluescript plasmid (Stratagene, Inc.) contains the N terminus of the lacZ gene, the recombinant β-galactosidase fusion protein can be expressed by IPTG induction of the lac promoter upstream from the lac gene. The fusion protein from tek-B1 was recognized by both a monoclonal antibody (#52-3-3, W. Steffen and R. Linck, unpublished observations) and a polyclonal antibody (Linck et al., 1987), both specific for sperm flagellar tektin B (Fig. 2). Compared to the apparent mass of flagellar tektin B (∼51 kDa) measured by SDS-PAGE, a major immunoreactive fusion polypeptide of ∼57 kDa was observed, along with two minor bands, which are presumably degradation prod-ucts of the fusion protein; degradation of fusion proteins in E. coli systems is commonly observed (Sambrook et al., 1989). The increased size of the fusion protein compared to the apparent mass of native tektin B subunit is partially due to the addition of the 36 amino acid N-terminal pep-tide of β-galactosidase (see further discussion below).

Clones tek-B1, tek-B2 and tek-B3 were sequenced; only that of tek-B1 is presented here (Fig. 3). Compared to the predicted coding region of tek-B1, tek-B2 is identical and tek-B3 differs by only 1 bp (A replaces G at position 960), which does not change the predicted amino acid. tek-B1 is 2,621 nucleotide in length. A stretch of 41 adenine bases defines the 3′ end, and a consensus polyadenylation sequence, AATAAA (Berget, 1984), begins at nucleotide 2,562.

The predicted amino acid sequence of the longest open reading frame beginning with the first methionine in tek-B1 is presented in Fig. 4. Proof that tek-B1 is the cDNA for the ciliary tektin B polypeptide was obtained by com-paring the amino acid sequence predicted from the cDNA with actual, partial amino acid sequence of proteolytic frag-ments of flagellar tektin B1. (Note: the ciliary and flagel-lar forms of tektin B are immunologically similar, have the same apparent molecular mass, and thus are thought to be homologous; Linck et al., 1987). Flagellar tektin B was purified to homogeneity by RP-HPLC and cleaved by cyanogen bromide. The CNBr-generated fragments were resolved by RP-HPLC, and peak fractions were collected. Three fragments were sequenced by automated Edman degradation (Table 1). There is close correspondence between the sequences predicted from the tek-B1 cDNA and the sequences of these three fragments (totaling 100 residues). The few discrepancies could be due to actual dif-ferences between ciliary and flagellar forms of tektin B or to occasional ambiguities in the amino acid sequence analy-sis. Thus, clone tek-B1 codes for tektin B and not for inter-mediate filament proteins (IFP, including nuclear lamins) with which anti-tektin antibodies have been shown to cross-react (Chang and Piperno, 1987; Norrander et al., 1992; Steffen and Linck, 1989a).

Translation beginning with the first ATG (AUG) as the initiation site would produce a ciliary tektin B polypeptide with a relative molecular mass of 46,260, which is ∼10% smaller than the 51,000 for flagellar tektin B, as estimated from SDS-PAGE (Linck and Langevin, 1982). This discrepancy may be due to tektin B running with an anom-alously high relative molecular mass in SDS-PAGE, since the tektin B fusion protein, including the 36 residues of β-galactosidase (∼3.9 kDa) and the 34 residues translated before the first Met in Fig. 3, has a predicted relative mol-ecular mass of 53,900 but migrates as a ∼57 kDa protein. This situation is similar to tektin A, which behaves as a ∼55 kDa protein in SDS-PAGE, though a relative molecular mass of 52,955 is predicted from its cDNA (Norrander et al., 1992). Alternatively, tek-B1 may be lacking a short N-terminal piece of the tektin B sequence. In this case, the 34 residues possibly encoded by bases 2-103 of Fig. 3 would merely increase the size of the non-helical N-termi-nal domain. Since there would be no change in the predicted rod domain, our structural analysis of tektin B and comparison with tektin A would not be affected (see Discussion).

As summarized in the Introduction, available evidence indicates that tektins A, B and C exist as extended, fibrous polymers associated with a particularly stable segment of the A-microtubule wall, i.e. the Sarkosyl-insoluble ribbon of protofilaments. We are ultimately interested in the function of tektins in microtubules. As a first step, to understand the potential interactions among the tektins and between tektins and tubulin, we have analyzed the primary and secondary structure of the deduced amino acid sequence of tektin B1, and we have compared the structures of tektins A and B.

Our analysis of the predicted secondary structures of the two proteins is summarized in Fig. 5, which suggests that the tektin A and B polypeptide chains each consist of several α-helical coiled-coil regions connected by spans of non-helical linkers. Predictions of α-helix (see Materials and Methods) are shown at the top of each panel in Fig. 5(a) and (b). Breaks in the helical structure are expected where each curve dips below the zero line. The potential for α-helical coiled-coil formation was determined by the method of Lupas et al. (1991), as shown in the lower plots in Fig. 5(a) and (b). This method scores each stretch of sequence according to its ability to form an α-helix with non-polar amino acids lined up along a slowly winding helical surface, which would allow a pair of polypeptides to associate as a coiled-coil. By this method the four major potential α-helical stretches of each tektin are predicted to form a coiled-coil structure (Fig. 5(c)), though there are a few intervening short potential α-helices that may not associate in the same way. In particular, predicted helices at the start of both segment 1 and segment 2 in tektin A (residues 87-115 and 279-298, respectively) and at the start of segment 2 in tektin B (residues 190-205) have too high a ratio of charged to hydrophobic residues for hydrophobic forces to hold identical regions together. (The possible significance of this is discussed below.) The four major regions of coiled-coil predicted for both tektins A and B are in simi-lar relative positions, so already there would appear to be a structural homology between tektins A and B, and indeed they do show significant primary sequence homology (see below). The coiled-coil models for tektins are consistent with the observed ∼70% combined α-helical content of tek-tins A and B (Linck and Langevin, 1982) and the strong α-type X-ray patterns of tektin filaments (Beese, 1984).

The results of a search for internal homology within tektin B1 using the DIAGON program (Staden, 1982; see Ma-terials and Methods) is shown in Fig. 6(a). This analysis, in which a span of 11 residues was used to assess short-range similarities and a span of 49 residues (approximately the size of an individual section of coiled-coil) for longer-range homologies, seems to indicate that the first and second halves (segments 1 and 2) of the polypeptide chain are only weakly homologous, except for two short runs, i.e.

DEEIRKLE vs EEEIAKQE and ERALEAKALPLDVA vs EQAIKDKENPLKLA.

Indeed, only an average of 22% of the residues are iden-tical when long central stretches selected from each half are compared (see Fig. 7). Nevertheless, such levels of homol-ogy must be significant when maintained over a long dis-tance (i.e. 27 identities plus 18 highly similar residues in a pair of 124-residue stretches with no inserted gaps). Oddly, the comparison of tektin B1 with tektin A1 (see below) pro-vides stronger evidence for the homology between the two halves of tektin B1.

Using the DIAGON program as above, we also searched for sequence homologies between tektins A1 and B1, both for the short-range 11-residue span (Fig. 6(b)) and for the long-range 49-residue span (Fig. 6(c)). By this analysis there is a high probability of short-range and long-range homology between tektin A (residues 113-423) and B (residues 32-344). Furthermore, there is significant homology between segment 1 of tektin A and segment 2 of tektin B, and between segment 1 of tektin B and segment 2 of tektin A. This analysis strongly suggests that sea urchin tektins A and B have a common evolutionary origin; it also suggests that tektins evolved from an early ancestral, half-length gene, which duplicated to produce the full-length sea urchin tektin genes. Since there is more homology between heterologous segments of tektins A and B than between segments within tektin B, it would appear that the gene duplication event preceded the divergence of tektins A and B. Investigations of tektins from evolutionarily simpler organisms should help to confirm this idea.

The sequence homology between tektins A and B is examined at the level of amino acid residues in Fig. 7. Given the homology between tektin segments indicated by the DIAGON plots, Fig. 7 shows segments 1 of tektins A and B aligned over segments 2 of tektins A and B. In this format one can more easily appreciate the amino acid iden-tities (34%) and conservative substitutions (20%) between tektins A and B within a span of 380 residues. The homol-ogy between tektins A and B in segment 2 (residues 353-403 of tektin A) is particularly striking, with 36 identities in 51 residues. Within this region the longest run of identities is RPNVELCRD.

Except for its homology with tektin A1, tektin B1 has no high degree of homology with other protein sequences available through EMBL and GenBank as of June 1993. At most, tektin B shares ∼13% sequence identity with inter-mediate filament proteins (IFP). We have previously noted a number of biochemical and immunological similarities between tektins and the IFP (see Introduction). At the level of primary sequence, tektin A1 was found to be distinctly different from the IFP superfamily (Norrander et al., 1992). First, tektin A1 lacks the IFP consensus sequence LNDRL(or F)AXYI at the start of the IFP rod domain 1A. Tektin B1 also lacks this sequence. Tektin A1 has only part of the IFP consensus sequence LD(or E)XEIAXYRKL-LEGEEXR(or K) at the C terminus, lacking the LLEGE-EXR motif. Within this region tektin B1 has no significant homology with either IFP or tektin A. Yet, despite these major differences, tektins A and B appear to have an over-all structural pattern rather similar to IFP (see Fig. 5c); i.e. both tektins and IFP are predicted to have a central rod composed of helical segments joined by non-helical link-ers, and the length and positions of these domains are approximately equal (compare Norrander et al., 1992, with Conway and Parry, 1988, and Steinert and Roop, 1988; Stewart, 1990, 1993). Thus, tektins are a class of proteins distinct from IFP, but the possibility remains that tektins and IFP are related by a common distant ancestor. Verification of such an evolutionary relationship might come from sequence studies of tektins and nuclear lamins in evolutionarily more primitive organisms.

The mechanisms by which tektins associate to form extended polymers and interact with tubulin are not known. Filament preparations can be obtained that are composed of equimolar amounts of tektins A and B only (unpublished observations), with the possibility of these filaments being either homopolymers or heteropolymers. The only (cir-cumstantial) evidence available is consistent with a het-eropolymer model. First, it has so far not been possible to separate tektins A and B without solubilizing both subunits. Second, tektins A and B have significantly different pI values, i.e. ∼6.9 and ∼6.2, respectively. In these respects, tektins would appear more similar to keratin types I and II, which are obligate heteropolymers, than to IF protein type III (e.g. desmin, GFAP and vimentin), which form homopolymers (Albers and Fuchs, 1992; Coulombe and Fuchs, 1990, Hatzfeld and Weber, 1990; Quinlan et al., 1986; Steinert, 1990). If and how tektin C is associated with the core filaments of tektins A and B are not understood at this time.

The basic units of tektin polymers are assumed to be polypeptide dimers, which could be homo-or heterodimers, given the close structural similarity predicted for the two polypeptides (Fig. 5c). In either case, we assume that two polypeptides assemble in parallel to form a polar dimer, as is the case for all known coiled-coil structural proteins (Lupas et al., 1991). Tektins differ from IF proteins in the center of their rod domains by having minor helices, which apparently lack sufficient hydrophobic residues to bond them. However, it is possible that if certain negatively charged residues in positions a or d of a heptapeptide on one strand were associated with positively charged residues in positions a or d on the other strand, then these sections of rod might be held together by ionic bonds.

This possibility is explored in Fig. 8. In tektin B, the arrangement of charged residues in heptad positions a and d at the beginning of segment 2 would bring most of them close to an opposite charge on the neighboring strand in a homodimer. However, Asp224 (marked by a double aster-isk), which would lie next to Glu228 on an α-helix, would not be counter-balanced in a homodimer. In a heterodimer, Asp224 of tektin B might bond to Arg309 of tektin A. Sim-ilarly, Glu284 of tektin A might be compensated for by Arg201 of tektin B, and Asp232 of tektin A by Arg142 of tektin B. Thus, it is possible that tektin AB heterodimers might be more stable than either type of homodimer. Crosslinking studies of tektin oligomers, which are under-way, should answer this question.

While tektins probably interact with tubulin (see Intro-duction), direct evidence is lacking. By way of comparison with the interactions of tropomyosin and actin, a 14-fold periodicity due to the arrangement of negatively charged and non-polar amino acids is present in tropomyosin, which interacts with ∼7 actin subunits along the actin filament (McLachlan and Stewart, 1976). However, neither tektin A (Norrander et al., 1992) nor tektin B has any obvious peri-odicities, apart from their 1 nm non-polar heptapeptide repeats, that would indicate potential interactions with tubulin monomers or dimers, with axial spacings of 4 and 8 nm, respectively. If they are linear molecules, tektin dimers, whether homo-or heterodimer, would measure nearly 50 nm long, a distance that could span six tubulin dimers or 48 nm along a protofilament. The strong homology between segments 1 and 2 further creates the possibility of a ∼24-nm spacing in tektin filaments. It may worth noting that the approximate 24 and 48 nm spacings potentially associated with tektins are multiples of the tubulin dimer repeat and correspond to spacings of dynein arms and subspacings of radial spokes (Amos and Klug, 1974; Darnell et al., 1986). Current studies are designed to elucidate the mechanism of tektin polymerization and to investigate whether tektins interact with tubulin and possibly with other axonemal com-ponents.

This work was supported by US Public Health Service grant GM35648 and Minnesota Medical Foundation grant SMF-527-86 to R.W.L. We thank Dr Eric Eccleston and Mr Bruce McKenzie in the Microchemical Facility at the University of Minnesota for the amino acid sequencing, and Mr Dave Lee for facilitating our work within the Department of Cell Biology and Neuroanatomy. The sequence reported in this paper has been deposited in the Gen-Bank data base (accession no. L21838).