Structural Comparison of Tektins and Evidence for Their Determination of Complex Spacings in Flagellar Microtubules

J. Mol. Biol. (1996) 257, 385–397

Structural Comparison of Tektins and Evidence for Their Determination of Complex Spacings in Flagellar Microtubules Jan M. Norrander1, Catherine A. Perrone1, Linda A. Amos2 and Richard W. Linck1 1

Department of Cell Biology and Neuroanatomy University of Minnesota 321 Church St., Minneapolis MN 55455, USA 2 Laboratory of Molecular Biology, Medical Research Council, Hills Road Cambridge, CB2 2QH, UK

Recent structural studies indicate that a tektin heteropolymer forms a unique protofilament of flagellar microtubules. We report here the sequence of tektin C (047 kDa), predicted from its cDNA (GenBank U38523), compared to tektins A (053 kDa) and B (051 kDa) from sea urchin (Strongylocentrotus purpuratus) sperm flagellar microtubules, and compared to partial sequences reported from mouse and human. We are now able to make several observations concerning the tektin family: (1) their common structural features, (2) a comparison of their structure to intermediate filament proteins, and (3) their possible organization in the tektin filament polymer. The predicted amino acid sequence identities/ similarities are: for tektins A and C, 42/54%; for tektins A and B, 34/51%; for tektins B and C, 29/42%; for tektin C and a partial cDNA clone from mouse testis, 55/65%; and for tektin B and a partial cDNA clone from human brain, 45/47%. The three tektins (and the human clone) possess the exact sequence repeat RPNVELCRD. The structural pattern of all three tektin polypeptides is similar to intermediate filament proteins. Tektins are predicted to form extended rods composed of two a-helical segments (0180 residues long) capable of forming coiled coils, which are interrupted by short non-helical linkers. The two segments are homologous in sequence and secondary structure, indicating a gene duplication event prior to the divergence of the three tektins. Along each tektin rod cysteine residues occur with a periodicity of approximately 8 nm, coincident with the axial repeat of tubulin dimers in microtubules. From EM data and calculations of secondary structure, the segment length of tektin AB heterodimers is likely to be 16 nm. Both segments of tektin C may be 24 nm long, but one may be 16 nm. On the basis of the available evidence, we propose that coassembly of tektin AB heterodimers with tektin C dimers produces filaments with overall repeats of 8, 16, 24, 32, 40, 48 and 96 nm, generating the basis for the complex spatial arrangements of axonemal components. 7 1996 Academic Press Limited

Keywords: centriole; coiled coil; dynein; intermediate filament; tubulin

Introduction Microtubules perform numerous, essential cellular functions in which the structural properties of microtubules are critical (Amos & Amos, 1991; Dustin, 1984). However, certain mechanisms of microtubule structure and function remain elusive. Abbreviations used: pf, protofilament; IF, intermediate filament; IFP, intermediate filament proteins. 0022–2836/96/120385–13 $18.00/0

For example, the mechanisms are not known that determine the three-dimensional form and stability of doublet and triplet microtubules of cilia, flagella, basal bodies and centrioles. Also not understood are the factors that determine the complex organization of components periodically arranged along microtubules or asymmetrically positioned around them, such as dynein arms, radial spokes and nexin links. We have been studying a family of proteins, called tektins, originally identified in sea urchin sperm flagellar microtubules, and probably ubiquitous, 7 1996 Academic Press Limited

386 that are relevant to these questions of microtubule structure and function. Three tektins have so far been characterized extensively from the sea urchin Strongylocentrotus purpuratus, tektin A 053 kDa, B 051 kDa and C 047 kDa, which originate from a specialized region of axonemal microtubules (Linck et al., 1985; Linck & Langevin, 1982; Linck & Stephens, 1987). Flagellar axonemes from sea urchin sperm (Linck, 1976; Meza et al., 1972) and from the alga Chlamydomonas (Witman et al., 1972) can be fractionated with Sarkosyl detergent into stable ribbons of 3-4 protofilaments (pf). The pf-ribbons originate from a specific region of the doublet microtubules and apparently also central pair singlet microtubules (Linck, 1976, 1990; Stephens et al., 1989). In sea urchin doublet microtubules the pf-ribbons are composed principally of a- and b-tubulin, tektins A, B and C, 77 and 83 kDa polypeptides, and several smaller polypeptides (Linck, 1976; Linck & Langevin, 1982). Antibodies raised against tektins stain centrioles (including human), basal bodies and the lengths of ciliary and flagellar doublet microtubules, and label 2 to 5 nm diameter fibrils extending out the ends of isolated pf-ribbons (Amos et al., 1985, 1986; Linck et al., 1985; Steffen & Linck, 1988; Steffen et al., 1994). Crosslinking studies (Pirner & Linck, 1994) have shown that tektins exist within the pf-ribbons as longitudinally oriented, heterodimeric polymers with axial periodicities matching the tubulin lattice. Recent structural studies have shown that one of the protofilaments of the ribbon is not composed of tubulin and concludes that it must instead be the tektin polymer (Nojima et al., 1995). Of further intrigue are the biochemical, structural and immunological similarities that tektins share with the intermediate filament (IF) protein superfamily (cf Fuchs & Weber, 1994; Norrander et al., 1992). Polyclonal and monoclonal antibodies raised against tektins crossreact strongly with numerous IF networks, nuclear envelopes, and IF proteins (Chang & Piperno, 1987; Steffen & Linck, 1989a,b). Similar to keratins, tektins are 070% a-helix (Linck & Langevin, 1982), yield strong a-type X-ray diffraction patterns (Beese, 1984), and are heteropolymers (Pirner & Linck, 1994). By cDNA sequence analysis (Chen et al., 1993; Norrander et al., 1992) each tektin polypeptide chain consists of a central rod of 0360 amino acids composed of several regions strongly predicted to form coiled coils, similar to IF proteins (Fuchs & Weber, 1994; Stewart, 1993). Yet, there are sufficient differences to indicate that tektins are a distinct family of coiled-coil proteins. While most of the structural chemistry on tektins has been done on sperm flagella, our work here is a molecular analysis of ciliary tektins, which are expressed during ciliogenesis of blastula stage sea urchin embryos (Amos et al., 1985; Norrander et al., 1995; Stephens, 1977, 1989; Stephens et al., 1989). We have previously reported the primary structure of ciliary tektins A1 and B1 (Norrander et al., 1992;

Tektin Structure

Figure 1. Analysis of sea urchin embryonic mRNA probed with tekC9-3 cDNA. Cytoplasmic RNA (lane a) and poly(A)+ mRNA (lane b) from blastula stage embryos of S. purpuratus were resolved electrophoretically. Filters were then probed with 32P-labeled nick-translated tekC9-3, which hybridizes with two mRNA bands of 2400 and 2650 nucleotides. The pBluescript SK plasmid alone did not hybridize to sea urchin mRNA (not shown). Molecular weights were determined by calibration with l DNA digested with EcoRI and HindIII.

Chen et al., 1993), deduced from cDNAs from a lZapII S. purpuratus blastula expression library. Here we analyze the predicted sequence and protein structure of tektin C1. We compare the three different tektins to each other, to recently reported cDNA sequences from mouse (Yuan et al., 1995) and human, and to IF proteins. Finally, given the predicted structures of tektins, we suggest principles that might govern aspects of microtubule structure and function.

Results Using affinity-purified, polyclonal antibodies raised against S. purpuratus sperm flagellar tektin C, we screened the lZapII sea urchin blastula expression library, isolating six putative tektin C clones. The size of the cDNA inserts contained in these clones was determined by digestion of isolated DNA with EcoRI, followed by resolution of the resulting fragments on an agarose gel. Four clones (tekC1-3, tekC9-3, tekC10-15 and tekC14-14) contained similar sized inserts (2300 to 2600 bp); the remaining two clones contained inserts of approximately 1900 bp. The 5' ends of the four largest clones were sequenced, and tekC9-3 was found to

Figure 2. Comparison of the predicted amino acid sequences of cDNAs for sea urchin tektins A1 (A), B1 (B) and C1 (C), and partial cDNAs from mouse (M) testis (MTEST638; Yuan et al., 1995) and from fetal human (H) brain (see Hillier et al., 1995). Asterisks are positioned between identical residues of tektins, and lines between residues with similar charge or hydrophobicity; the mouse clone best compares to tektin C, and the human clone best compares to tektin B. Underlined sequences are predicted to form a-helices. The identities/similarities (identical plus similar residues) between pairs of tektins are respectively: 34%/51% for tektins A1 and B1; 42%/54% for tektins A1 and C1; and 29%/42% for tektins B1 and C1. The sequence repeat RPNVELCRD (referenced to tektin A at residues 378 to 386) is present in all three tektins and in the human clone. Four conserved cysteine residues occur in two or more tektins, referenced to tektin C at residues 115, 188, 329, and 390, and three conserved tryptophan residues occur in two or more tektins, referenced to tektin C residues 16, 73, and 217. GenBank data base accession numbers are: M97188 for S. purpuratus tektin A1; L21838 for tektin B1; U38523 for tektin C1 (clone tekC9-3); Z31216 for the mouse MTEST638 clone; T78294 for the human clone.

388 contain an additional 142 bp of sequence on its 5' end compared with tekC1-3, tekC10-15 and tekC14-14. As a result, tekC9-3 was chosen for further study. On a blot of S. purpuratus blastula mRNA, tekC9-3 hybridizes to two bands of approximately 2650 and 2400 bases (Figure 1). These mRNAs are specifically up-regulated during ciliogenesis in several sea urchin species (Norrander et al., 1995). Clone tekC9-3 was sequenced, giving a cDNA of 2563 base-pairs. The longest open reading frame predicts a 402 amino acid protein with a molecular weight of 46,567, referred to here as tektin C1 (Figure 2). Clone tekC9-3 contains 170 bases of untranslated 5' sequence, including an in-frame stop codon upstream from the predicted initiator methionine, and 1187 bases of 3' untranslated sequence. The identity of tekC9-3 as a cDNA clone for tektin C is supported by several criteria. First, the predicted molecular weight of 46,567 for the tektin C1 protein compares favorably with the SDS-PAGE estimate of 47,000 for tektin C (Linck & Langevin, 1982; Linck & Stephens, 1987; Linck et al., 1987). Most directly, the predicted amino acid sequence of ciliary tektin C1 shows an identical match with 57 residues obtained from proteolytic fragments of flagellar tektin C (Table 1); the predicted amino acid composition of tektin C1 also compares favorably with experimentally determined values of tektin C (Linck & Stephens, 1987). Finally, tektin C1 is homologous to, but distinct from tektins A1 and B1 (Figure 2). Importantly, the sequence comparison of tektins A1, B1 and C1, and a reported human brain EST clone (Hillier et al., 1995), reveals the exact sequence repeat RPNVELCRD. We analyzed the amino acid sequence in various standard ways, as shown in the following sets of figures. Sequence homologies were sought both within tektin C1 and between tektin C1 and tektins A1 and B1 (Figure 3 to 5, and Table 2). Secondary structure prediction algorithms were applied to the sequence and the results compared with those for other tektins and for intermediate filament proteins (Figure 4). The distributions of different classes of residues in the sequences of tektins A1, B1 and C1 were also investigated, to discover any local concentrations (as in Figure 4(d)) or periodicities in the arrangements of charged and hydrophobic residues (Figure 6). The results presented in the Figures are further discussed below.

Analysis and Discussion Homology of segments 1 and 2 in tektin polypeptide chains The amino acid sequences of sea urchin tektins A1, B1 and C1 have a close resemblance (Figures 2 and 5). For each pair of tektins the percent identities and percent similarities (identical residues plus conservative substitutions) are respectively: for tektins A1 and B1, 34% and 51%; for tektins A1 and

Tektin Structure

Table 1. Amino acid sequences of tektin C peptide fragments and of corresponding sequences predicted from the cDNAs Residues Peptide 1 cDNA Residues

T274-Q288 : D? R -TKGKLETHLAKVEGQ MTKGKLETHLAKVEGQM

Peptide 2 cDNA

T274-K306 : T -DKGKLETHLAKVEGQ-REMEENIQKLQKGVDDK MTKGKLETHLAKVEGQMREMEENIQKLQKGVDDKM

Residues Peptide 3 cDNA

E293-K306 : -EENIQKLQKGVDDK MEENIQKLQKGVDDKM

Residues Peptide 4 cDNA

T392-Y402 : -TLRRSINIKRY MTLRRSINIKRYstop

Residue numbers are from Figure 2. Peptide numbers refer to amino acid sequences of CNBr fragments obtained from the sperm flagellar tektin C protein. cDNA refers to the amino acid sequences deduced from matching regions of the tekC1 cDNA for the ciliary form of tektin C. Ambiguities or uncertainties in the peptide sequences are given above the main sequence line.

C1, 42%, 54%; and for tektins B1 and C1, 29%, 42%. These values are directly comparable to the relatedness of the tektins predicted from earlier peptide mapping studies (Linck & Stephens, 1987). In addition, a partial 98 residue sequence of a putative tektin from mouse testis (Yuan et al., 1995) is most homologous with tektin C1 (55%, 65%), and a 74 residue sequence from fetal human brain (Hillier et al., 1995) is homologous with tektin B1 (45%, 47%), see Figure 2. Within the sea urchin tektins the most conserved region spans tektin A1 residues 370 to 391 and contains a nine consecutive residue sequence repeat, R378 PNVELCRD386 , which is also present in the human clone. In addition, there are conserved cysteine residues and tryptophan residues, as described in Figure 2. Besides these direct sequence comparisons, tektins share a similar overall structural design (see below). Analysis of the amino acid sequence homologies within individual tektin polypeptide chains and between different tektin chains (Figure 3) suggests a relationship between the two halves (i.e. segments 1 and 2) of each tektin. For example, self-comparisons of tektin C1 to itself (Figure 3(g)) and tektin A1 to itself (Figure 3(h)) reveal significant internal homologies between segments 1 and 2. Similar homologies are also observed in pair-wise comparisons of different tektins (Figure 3(a) to (f)); e.g. there is a significant homology between segment 1 of one tektin and segment 2 of a different tektin. In all cases these homologies depend on many similar amino acids matching over a long range, although the short-range homology and the number of actual identities is fairly low (Figure 5 and Table 2). It is clear from Table 2 and by comparing Figure 3(g) (self-comparison of tektin C1) with similar plots for tektin A (Figure 3(h)) and tektin B (Chen et al., 1993), that tektin C has the most closely-related two halves, and that the two halves of tektin B are most

Tektin Structure

389 different, with the N-terminal half of the latter being least like any other. In general, however, equivalent halves (i.e. segments) of different proteins are more alike, than are different halves of the same protein, supporting the idea that the divergence into different tektin genes occurred after an initial duplication of a single, ancestral ‘‘halfgene’’. The six half-sequences are compared in detail in Figure 5 and Table 2. Of possible significance, two of the conserved cysteine residues (referenced to tektin C residues 188 and 390) line up in all six tektin half-sequences and in mouse MTEST638 (and so far, one occurs in the human clone), and two of the conserved tryptophan residues, (referenced to tektin C residues 16 and 217) line up in five of the six tektin half-sequences. Predicted secondary structure of tektins and their resemblance to IF proteins

Figure 3. DIAGON plots (Staden, 1982) showing amino acid sequence homologies between various pairs of tektins. 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 which corresponds to the level of similarity between sequence characters over a chosen sequence span centered on point x in sequence X and point y in sequence Y, allowing no insertions or deletions. Comparisons are as follows, by alignment of the N termini at the lower left corner of each panel: (a) and (b), tektins A1 versus C1; (c) and (d), tektins B1 versus C1; (e) and (f) tektins A1 versus B1; (g) tektin C1 to itself; (h) tetkin A1 to itself. (a), (c), (e) Show short-range homology scores (for an 11-residue span); a point here indicates a match with a probability of 6 × 10−4 of occurring randomly. (b), (d), (f) test for homologies that stay in register over a longer range (a 49-residue span); in this case the points represent higher stringency scores, with a probability of 10−5 of arising by chance. In (g) (self-comparison of tektin C1) and (h) (self-comparison of tektin A1) the points above and left of the diagonal are for short-range scores (as in (a), (c), (e)), while points below and to the right of the diagonal are for long-range scores (as in (b), (d), (f)). In all the plots, off-diagonal lines

Secondary structure predictions based on the amino acid sequences (Figure 4(b),(c)) confirm the close relationships between different tektins. In general, each tektin polypeptide chain is composed of several a-helical regions predicted to form coiled coils, and the coiled-coil regions of segment 1 correspond approximately to those in segment 2 of each tektin. Originally (Norrander et al., 1992), we adapted terminology from intermediate filament proteins (IFP) in referring to the different coiled-coil regions of segments 1 and 2, i.e. regions 1A, 1B(i), 1B(ii), and 2A, 2B(i) and 2B(ii); however, for tektins such terminology is confusing, and here we refer to the regions simply as 1a,b,c, and 2a,b,c (Figure 4). As indicated by the homology between segments 1 and 2 (Figure 3 and 5), the predicted structures of the different tektin polypeptide chains are particularly alike through regions 1b, 1c, 2b and 2c, but there is more variability in regions 1a and 2a. Especially in region 1a, tektin C1 has a strongly predicted coiled coil (i.e. continuous heptad repeats, where the first and fourth residues in the repeat are hydrophobic), whereas tektin Al has a-helix but lacks sufficient repeating residues to form coiled coil with itself. Tektin B1 may lack region 1a all together, but its reported N-terminal sequence may be incomplete due to a possible truncation of the cDNA (Chen et al., 1993). We have previously noted the structural similarity between tektins and IFP. Cytokeratins probably provide the most useful paradigm for this comparison. While type III IFP such as vimentin can form homopolymers, cytokeratins are obligate heteropolymers formed of type I (acidic) and type II (neutral-basic) polypeptides (Conway & Parry, 1988; Fuchs & Weber, 1994; Stewart, 1993), that are first arranged to form a parallel (N-terminal to N-terminal), coiled-coil heterodimer. Tetramers are (marked by asterisks) show the homologies between the first half (0segment 1) of one sequence and the second half (0segment 2) of the other sequence.

390

Tektin Structure

Table 2. Shared identities and similarities in tektin half-rods (as a measure of relatedness) 1st half of tektin A

B

2nd half of tektin

C

A

B

C

1st half of tektin A (187) 44 B 44 (152) C 77 48

77 48 (182)

33 30 36

37 31 41

35 32 45

2nd half of tektin A 33 30 B 37 31 C 35 32

36 41 45

(188) 78 78

78 (187) 65

78 65 (188)

Numbers of identical residues between two half-sequences when the tektin sequences are aligned as in Figure 5. These values show: (1) The same halves (segments) of different proteins are more similar than are different halves (segments 1 and 2) of the same protein, supporting the hypothesis of a gene duplication of an ancestral ‘‘half-gene/segment’’. (2) Tektin C has the most closely related two halves. (3) Tektin B has the least-related two halves; this is mainly because the first half of B is most different from everything else (and is not simply due to possible N-terminal truncation), and because the second half of B is slightly more different from the consensus in A or C.

arranged from anti-parallel dimers, which in turn are assembled in a staggered array to form the final 010 nm diameter intermediate filament (Fuchs & Weber, 1994). As with cytokeratins, tektins are highly a-helical (Beese, 1984; Linck & Langevin, 1982), tektins A and B have pI values of near-neutral (6.9) and acidic (6.2), respectively (Linck & Langevin, 1982), and they exist as a heterodimer (Pirner & Linck, 1994). Furthermore, the existence of a third component, tektin C, may make the tektin ABC relationship analogous to the neurofilament triplet NF-L, NF-M and NF-H (see below). As shown in our present study, the features of tektin C1 make it even more similar in overall structure to IFP (Figure 4, compare (c) and (e)), than is tektin A1, for which the comparison with IF proteins was originally made (Norrander et al., 1992). In making that comparison, we conjectured that the linker between rod segments 2b and 2c in tektins is equivalently positioned to the break in the heptad repeat pattern in the middle of segment 2 of IFP, and that the tektin linker between regions 1b and 1c is in the equivalent position of a 42-residue stretch of rod found in segment 1B of invertebrate IFP and all nuclear lamins but missing from vertebrate IFP (Weber et al., 1988, 1989). The additional structural information provided by the sequence of tektin C strengthens our original hypothesis that the two protein families (tektins and IFP) may have divergently evolved from a common ancestor, or have evolved convergently into surprisingly similar structures. However, significant features of tektins clearly distinguish them from the IFP family (Conway & Parry, 1988). First, tektins lack most of the consensus sequences of IFP (including LNDRL(or F)AXYI at the start of rod domain 1A, and LD(or E)XEIAXYRKLLEGEEXR(or K) at the end of rod 2B); see Figure 5. Second, the conserved tryptophan

residues of tektins are not in the position of the single W present in most IFP, including W312 of Ascaris lumbricoides (Fuchs & Weber, 1994). Third, the sequence repeat RPNVELCRD found in all known tektin sequences is absent in IF proteins. In addition, contrary to our initial expectations, none of the tektin sequences reveals a dramatic repeat of 28 or 56 residues (Figure 6) that in a coiled-coil rod segment would measure 4 or 8 nm, respectively, and would allow each tektin chain to interact directly with every tubulin monomer or dimer along a protofilament (Amos & Klug, 1974), as proposed for the giardins (Holberton et al., 1988). A 28-residue pattern of charged residues is a feature of all IF proteins (McLachlan & Stewart, 1982), although there is no evidence that their rod segments interact closely with microtubules. In this case, the arrangement of charges is presumably needed mainly for forming the 10 nm bundles of coiled coils; whereas tektin filaments may instead be held together by other forces (e.g. disulfide bridges; see below) and grow to a smaller diameter more equivalent to an IF protofilament (2–3 nm) or protofibril (4.5 nm). The more complex arrangement of charged residues in tektins is probably related to their special functions (see below). Principles of subunit associations in tektin filament assembly The associations of tektin subunits in the assembled filament and in pf-ribbons can be interpreted to some degree on the basis of the present sequence analysis and established biochemical parameters. Within the pf-ribbon, derived from near the inner A–B–tubule junction, tektins form a continuous longitudinal filament and remain as such after crosslinking with N-hydroxy-succinimidyl esters and subsequent extraction of the ribbons with urea (Pirner & Linck, 1994). Nojima et al. (1995) have shown that each ribbon includes a distinct pf that is not tubulin; in the isolated ribbon this unique pf has less mass than a tubulin pf and fails to bind kinesin. Together, the evidence indicates that this unique pf is the tektin filament, and that it lies directly between two tubulin pfs, rather than being closely associated with a single tubulin pf. Finally, isolated filament preparations are composed of approximately equimolar amounts of tektins A, B and C; however, tektin C can be selectively solubilized, leaving filaments composed of equimolar tektins A and B, which can be crosslinked as AB-heterodimers (Pirner & Linck, 1994). This evidence would suggest that the tektin filament is built of tektin AB heterodimer coiled coils to form a core filament with tektin C added peripherally. Although it is not yet known whether the subunits of the AB dimer are parallel (as with IFP) or anti-parallel, all natural coiled coils are parallel (Stewart & McLachlan, 1975). From the sequence analysis we note that if tektins A and B are arranged in parallel (Figure 2, 5 and 7i), then any

Tektin Structure

391 Figure 4. Predictions of the proteins’ secondary structure from their amino acid sequences. (a) A model summarizing the predictions for tektin C1: open boxes stand for a-helix and shading indicates the presence of a seam of hydrophobic residues along one side, which would stabilize pairs of helices in the form of a coiled-coil homodimers, as shown. (b) Plots from ANALYSEP (Staden, 1988) for tektins C, A and B of the probability that any point in these sequences will form a-helix, according to the criteria of Robson et al. (1976). (c) Plots from NEWCOILS (Lupas et al., 1992) for tektins C, A and B of the probability that segments of sequence containing heptad repeats will form coiled coil (obtained using a 28-residue span). The predicted segments of coiled coil are labeled 1a,b,c and 2a,b,c. (d) Distribution of charge (averaged over a 11-residue span) in tektins C, A and B (basic/ positive charge above the horizontal line, acidic/negative below). If the sign of tektin A segment 1 is reversed (*), it matches with segment 2 of tektin B, indicating a possible charge interaction between adjacent heterodimers (see Figure 7ii; and Pirner & Linck, 1994). (e) Plots from NEWCOILS for three representative IFP, namely fly lamin c from Drosophila (Bossie & Sanders, 1993), and IFP from the roman snail Helix pomatia (Weber et al., 1988) and the nematode worm C. elegans (Weber et al., 1989). The broken lines indicate a break inserted into the sequence of segment 2, at the site of a dislocation in the hydrophobic seam. All sequences are depicted to scale.

one of four aligned pairs of cysteine residues could form interchain disulfides between the non-helical linker regions that might loop out from the coiled-coil backbone (compare tektin A residues 173, 243, 384, 445, and tektin B residues 89, 162, 303, 363). Interchain disulfides within a coiled coil and/or between adjacent coiled coils could explain why heterodimers are observed in SDS in the absence of reducing agents (Pirner & Linck, 1994); the cysteines may have other functions as well (see below). Although tektin tetramers have not specifically been observed, tektins A and B do form short oligomers (0600 kDa) and can assemble without tektin C to form extended filaments similar in appearance to the filaments isolated from microtubules (unpublished observations). Like tektin B, tektin C has an acidic isoelectric point (pI 6.15), but many of its excess

acidic amino acids are arranged to fit in with the heptapeptide repeat (on the surface of a coiled coil), which would allow for the formation of homodimers. The results of Nojima et al. (1995) place rough limits on the mass of a tektin filament, ranging from 05 kDa/nm (for the residual fibrils appearing as natural breakdown products in specimens for STEM) to somewhat less than that of a tubulin pf (012 kDa/nm). Given the presence of roughly equimolar amounts of tektins A, B and C, and dimer lengths of 32 or 48 nm (see below), a tektin filament might consist of three or six dimer strands. For example, the model in Figure 7ii (consisting of 32 nm-long tektin AB dimers and 48 nm-long tektin CC dimers) would have a mass of 5 to 9 kDa/nm, depending on how much of the tektin polypeptides outside the 035 kDa rod

392

Tektin Structure

Figure 5. Separate halves of tektin sequences A1, B1 & C1 co-aligned to show the sequence homologies between segments 1(a,b,c) and 2(a,b,c), apparent from Figure 3. Asterisks lie between the two halves where the residues in five or six half-sequences are identical, vertical lines where the residues in five to six halves are homologous in charge or hydrophobicity. Underlined residues are predicted to form a-helix (cf Figure 4(b)). The letter C is shown above and below columns of conserved cysteine residues, and W above conserved tryptophan residues.

domains are closely associated with the tektin filament itself or with neighboring tubulin pfs. Three coiled coils (as in Figure 7iii) would be expected to pack together to give an overall diameter of about 4 nm, whereas a bundle of 6–7 coiled coils, as postulated in an earlier model (Linck & Langevin, 1982), would be much closer in size and mass to a tubulin pf. Tektin filaments have complex axial spacings The sequence analysis and direct structural observations indicate that the packing of tektin subunits in the assembled filament is likely to produce complex axial spacings. After bifunctional chemical crosslinking with DTSSP, the core tektin AB filaments possess small, repeating globular domains, spaced 16 nm apart (Pirner & Linck, 1994). These domains may represent a conformational change in the tektin subunits induced by the crosslinker (e.g. stretches of non-helical domains that are highly extended and therefore not normally visible by electron microscopy, but that become more compact after crosslinking). In earlier studies a 48 nm axial spacing was observed along tektin filaments (composed of tektins A, B and C) by labeling with a monoclonal antibody (Amos et al., 1986). We previously suggested that tektin AB heterodimers might be 48 nm long (Chen et al.,

1993), based on the maximum possible length of the a-helical rod and accounting for the observed 48 nm axial spacing. To combine these observations in a model (not shown), the tektin AB core filament might be composed of 48 nm subunits staggered by 16 nm, and to complete this model, tektin C homodimers might be added peripherally to generate additional periodicities. Alternatively, the tektin AB heterodimer might be only 32 nm in length, given that rod segment 1a may be missing from tektin B and that the Lupas et al. (1991) algorithm (see Figure 4(b)) does not predict coiled-coil structure for segment 1a of tektin A (unlike this region of tektin C or IFP) or for segment 2a in any of the three tektins. Rods formed only from segments 1b, 1c, 2b and 2c would measure approximately 32 nm long. A 16 nm repeat could then be explained, if the 32 nm long dimers assumed to form the core filament are half-staggered. Such an arrangement might allow the peak of excess acidic negative charges in segment lb of tektin A to match up with a peak of excess basic positive charges in segment 2b of tektin B in another heterodimer (as suggested in Figure 4(d)(A')), as well as linking up the proposed disulfides (see Figure 7i). Tektin C, on the other hand, may form homodimers approximately 40 nm long, which if assembled at intervals of 48 nm along the tektin AB filament core, would produce an overall axial

393

Tektin Structure

Interactions between tektin filaments and tubulin protofilaments There are strong reasons to expect a direct interaction between tektin and tubulin, including the close molar ratios of tektins and tubulin in the pf-ribbon and the apparently close structural relationship between a tektin filament and tubulin pfs (Nojima et al., 1995). The fact that reassociation studies have not yet yielded positive results may be due to conditions affecting in vitro reassociation (Pirner, 1995). As already mentioned, tektins do not have 28-residue/4-nm or 56-residue/8-nm repeating patterns of charged residues (Figure 6), as found in IFP and giardins. However, the final subunit arrangement in the tektin filament may generate equivalent 8 nm axial repeats. For example, we note that cysteine residues occur regularly along most of the tektin polypeptide chains, and that based on the secondary structures of the tektin rods, these cysteine residues would repeat at approximately 8 nm, i.e. the axial repeat of the tubulin dimer (Amos & Klug, 1974). It would be interesting if disulfide bridging provided the mechanism for tektin-tubulin interactions, since this could explain the stability of the 3–4 protofilament ribbons in Sarkosyl-detergent and chaotropic solvents. Predictions aside, the observed 16 nm periodicity along tektin AB filaments (Pirner & Linck, 1994) provides evidence for tektin filament interactions with every other tubulin dimer.

Structural implications for tektin functions

Figure 6. Fourier spectra obtained from the sequences of the rod domains of tektins C, A and B (minus the N and C-termini), according to the procedure described by McLachlan & Stewart (1982). In each case, positions occupied by a hydrophobic, acidic or basic residue were given values of one, while other positions were set to zero. In all three sequences there are peaks for hydrophobic residues corresponding to average spacings of 7/2 and 7/3 residues (the values in the Figure are reciprocals), thus indicating the presence of a hydrophobic seam along regions of a-helix. Tektin C also has a 7/1-residue peak for acidic residues, and tektin B has acidic peaks at 7/2 and 7/3 residues. Some weaker peaks indicate longer charged-residue periodicities of possible significance, including orders of possible 28 residue (04 nm) and 56 residue (08 nm) spacings.

periodicity of 96 nm in a complete tektin filament (Figure 7iii). These models are meant not as formal proposals but as illustrations of some of the principles by which tektin subunits might assemble to form filaments with complex axial spacings for purposes to be discussed below.

Although there are as yet no proven functions for tektin, their structural properties and location in flagellar microtubules support the previously suggested functions for this new class of proteins (Linck, 1990): Tektins may (1) stabilize tubulin protofilaments, (2) form the attachment points between A and B-tubules of ciliary/flagellar doublet microtubules and C-tubules of centrioles, and (3) generate complex, long-range patterns of binding sites for axonemal components, such as nexins, radial spokes and dynein arms. It seems unlikely to be just coincidence that the spacing of structures attached to flagellar microtubules (Amos et al., 1976) are closely related to the observed and predicted spacings within tektin filaments. In particular, the locations of structures proximally associated with the inner A–B–tubule junction, such as the radial spoke triplets (with 24/32/40 nm subspacings (in sea urchins) and an overall repeat of 96 nm), inner dynein arms (with an arrangement related to that of radial spokes; Mastronarde et al., 1992), and nexin links of 96 nm (Dallai & Bernini, 1973) may be generated directly or indirectly by sites on the tektin filament; a similar mechanism might account for the 16 and 32 nm spacings of components associated with central pair singlet

394

Tektin Structure

microtubules (Linck et al., 1981). Finally, the position of the tektin filaments and the supposed nexin links between them, forming a strong structural skeleton within the axoneme (Stephens et al., 1989), would allow them to provide anchor points for the main shear-producing and regulatory components, i.e. the inner dynein arms and radial spokes.

Materials and Methods Screening and cloning the cDNA library Sea urchins (Strongylocentrotus purpuratus) were obtained from Marinus, Inc., Long Beach, CA. A cDNA lZapII (Stratagene, Inc., La Jolla, CA) expression library was constructed previously from S. purpuratus blastula mRNA (Norrander et al., 1992). This library was screened with affinity-purified polyclonal antibodies raised against S. purpuratus sperm flagellar tektin C (Linck et al., 1987). Standard molecular biology protocols were used (Sam-

brook et al., 1989), as previously described (Chen et al., 1993; Norrander et al., 1992), and pBluescript plasmids containing tekC cDNA inserts were isolated from positive lZapII clones as described by Stratagene, Inc. (La Jolla, CA). RNA blots As previously reported (Norrander et al., 1992), 5 mg of sea urchin embryonic mRNA was resolved on 1.5% (w/v) agarose, 2.2 M formaldehyde gels (Sambrook et al., 1989), transferred to nitrocellulose (Thomas, 1980), and hybridized with radiolabeled tekC9-3, using the Nick Translation Kit from Gibco/BRL (Gaithersburg, MD). cDNA sequencing Overlapping, sequential deletion clones were constructed using the ExoIII/Mung Bean Nuclease Deletion Kit from Stratagene, Inc. Sequencing of single-stranded templates was performed according to the method of Sanger et al. (1977) using the Sequenase 2.0 Sequencing

Figure 7 legend opposite

Tektin Structure

Kit from United States Biochemical Corporation (Cleveland, OH). The cDNA sequence for clone tekC9-3 reported here, has been deposited in GenBank, accession number U38523. Protein purification and peptide sequencing Sperm flagellar tektins were purified by Sarkosyl/urea extraction and reversed-phase (RP) HPLC, as described by Linck & Stephens (1987). The purity of tektin C was assessed by SDS-PAGE. Tektin C was dissolved to 02.5 mg/ml in 0.4 ml 70% trifluoroacetic acid and treated with cyanogen bromide, at 030 mmol 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 peptides were collected and sequenced by automated Edman degradation on an Applied Biosystems 470 peptide sequencer, in the Microchemical Facility, University of Minnesota. Sequence analysis The sequence was analyzed using the program ANALYSEP (Staden, 1988), based on the studies of

395 Robson & Suzuki (1976), in order to predict stretches of sequence that are more likely to form a-helix than any alternative secondary structure. The potential for a-helical coiled-coil formation was determined by the method of Lupas et al. (1991), which scores any stretch of a-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 program DIAGON (Staden, 1982) with a proportional matching calculation, first described by McLachlan (1982). 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 probability of < 0.00001. In calculations of homologies and similarities, the following residues were considered to be conservative substitutions: D and E (acidic/negatively charged); K, H and R (basic/positively charged); F, I, L, M, V, Y (hydrophobic/nonpolar); S and T (possible phosphorylation sites).

Figure 7. Predicted structures of tektin AB heterodimer and tektin C homodimer and illustrations of their possible organization within a tektin filament. i, Each polypeptide chain is represented by open rectangles (predicted a-helical regions) and straight lines (nonhelical regions). Hatched areas between chains indicate predicted coiled-coil formation within the heterodimers/homodimers. The positions of sulfhydryls (S) at approximately 8 nm intervals are indicated. The interpretation of chain lengths is as follows: The tektin C dimer rod is almost entirely coiled coil (see Figure 4(c)), and thus, depending on the packing of the non-helical portions, the whole rod would measure 40 to 48 nm long. The C dimer may be divided into two approximately equal 24 nm segments, or into 24 and 16 nm segments, depending on whether the 2a domains form part of the rod or loop out from the rod. The situation with tektin AB heterodimers is more uncertain; however, tektin AB filaments are observed to have a 16 nm axial periodicity, possibly formed by a large protrusion from the filament axis (Pirner & Linck, 1994). A protruding feature could arise from some combination of the N-terminal domain and segments 1a and 2a, whose sequences fail to satisfy fully the requirements for a coiled-coil structure (see Figure 4(c)). The resulting AB dimer rod may then be only 32 nm long, divided into two equal 16 nm segments. ii, Possible arrangement of dimers in the assembly of a tektin ABC filament, based on an equimolar ratio of tektins A, B and C, and the fact that tektin C can be solubilized to leave tektin AB core filaments with a 16 nm axial periodicity (Pirner & Linck, 1994). Tektin AB heterodimers may be first assembled to form extended, half-staggered tetramers, with the excess negative/acidic charge of tektin A segment 1b interacting with the excess positive/basic charge of tektin B segment 2b. The 08 nm repeating sulfhydryls (*) may act to stabilize the dimers and the filament, but it may also be that these groups form disulfides with the 8 nm repeating tubulin dimers in adjoining protofilaments. The core AB filaments would show a strong 16 nm periodicity, and the overall repeat would depend on the coiled-coil pitch (see Figure 7iii). Next, tektin C homodimers are assembled with a 48 nm periodicity to account for the observations of this repeat by EM (Amos et al., 1986; Nojima et al., 1995); an overall axial repeat of 96 nm occurs as the lowest common denominator of the 32 nm and 48 nm periodicities. iii, Possible relationship of three coiled-coils, as seen from the side and in successive cross-sections (note the differences in cross-sections 16 nm apart, which is why the overall periodicity is 96 nm, not 48 nm). For simplicity, the tektin rod domains are shown with only the N-terminal projections from tektins A and C. Since the tektin filament has been shown to lie between two tubulin protofilaments (Nojima et al., 1995), it could be formed by a simple, untwisted bundle of three coiled-coil rods (see the text for discussion of filament mass), able to interact in a similar way with the 8 nm periodic tubulin dimer lattice at intervals of 8, 16, 32 or 48 nm; if instead, the coiled-coils were themselves twisted into a higher order helical arrangement, the axial repeat would be much longer and the interactions with tubulin very varied. A simple arrangement is possible if the coiled-coil pitch is 16, 32 or 48 nm; of these, 32 nm, the proposed length of the tektin AB heterodimer, would allow charges (Figure 7ii) to match at 16 nm intervals. The model requires a polar arrangement of tektin AB heterodimers. The CC homodimers are arbitrarily shown in the same orientation.

396

Acknowledgements This work was supported by US Public Health Service grant GM35648, Minnesota Medical Foundation grants SMF-527-86 and FSW-155-95 to R.W.L., by National Research Service Award 1F32-GM11669 to J.M.N., and by National Science Foundation Research Training Group grant BIR 9113444. We thank the reviewers for suggestions to improve the original manuscript.

References Amos, L. A. & Amos, W. B. (1991). Molecules of the Cytoskeleton, p. 253, Guilford Press, New York, NY. Amos, L. A. & Klug, A. (1974). Arrangement of subunits in flagellar microtubules. J. Cell Sci. 14, 523–549. Amos, L. A., Linck, R. W. & Klug, A. (1976). Molecular structure of flagellar microtubules. In Cell Motility (Goldman, R., Pollard, T. & Rosenbaum, J., eds), pp. 847–867, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Amos, W. B., Amos, L. A. & Linck, R. W. (1985). Proteins closely similar to flagellar tektins are detected in cilia but not in cytoplasmic microtubules. Cell Motil. 5, 239–249. Amos, W. B., Amos, L. A. & Linck, R. W. (1986). Studies of tektin filaments from flagellar microtubules by immunoelectron microscopy. J. Cell Sci. Suppl. 5, 55–68. Beese, L. (1984). Microtubule structure: an analysis by X-ray diffraction. Ph.D. thesis, Brandeis University, Waltham, MA. Bossie, C. A. & Sanders, M. M. (1993). A cDNA from Drosophila melanogaster encodes a lamin C-like intermediate filament protein. J. Cell Sci. 104, 1263–1272. Chang, X.-j. & Piperno, G. (1987). Cross-reactivity of antibodies for flagellar tektin and intermediate filament subunits. J. Cell Biol. 104, 1563–1568. Chen, R., Perrone, C. A., Amos, L. A. & Linck, R. W. (1993). Tektin B1 from ciliary microtubules: primary structure as deduced from the cDNA sequence and comparison with tektin A1. J. Cell Sci. 106, 909–918. Conway, J. F. & Parry, D. A. D. (1988). Intermediate filament structure: 3. Analysis of sequence homologies. Int. J. Biol. Macromolecules, 10, 79–98. Dallai, R. & Bernini, F. F. G. (1973). Interdoublet connections in the sperm flagella. J. Submic. Cytol. 5, 137–145. Dustin, P. (1984). Microtubules, p. 482, Springer-Verlag, New York. Fuchs, E. & Weber, K. (1994). Intermediate filaments— structure, dynamics, function, and disease. Annu. Rev. Biochem. 632, 345–382. Hillier, L., Clark, N., Dubuque, T., Elliston, K., Hawkins, M., Holman, M., Hultman, M., Kucaba, T., Le, M., Lennon, G., Marra, M., Parsons, J., Rifkin, L., Rohlfing, T., Soares, M., Tan, F., Trevaskis, E., Waterson, R., Williamson, A., Wohldmann, P. & Wilson, R. (1995). The WashU-Merck EST Project. GenBank accession number T78294. Holberton, D., Baker, D. A. & Marshall, J. (1988). Segmented a-helical coiled-coil structure of the protein giardin from the Giardia cytoskeleton. J. Mol. Biol. 204, 789–795. Linck, R. W. (1976). Flagellar doublet microtubules: fractionation of minor components and a-tubulin from specific regions of the A-tubule. J. Cell Sci. 20, 405–439.

Tektin Structure

Linck, R. W. (1990). Tektins and microtubules. Advan. Cell Biol. 32, 35–63. Linck, R. W. & Langevin, G. L. (1982). Structure and chemical composition of insoluble filamentous components of sperm flagellar microtubules. J. Cell Sci. 58, 1–22. Linck, R. W. & Stephens, R. E. (1987). Biochemical characterization of tektins from sperm flagellar doublet microtubules. J. Cell Biol. 104, 1069–1075. Linck, R. W., Olson, G. E. & Langevin, G. L. (1981). Arrangement of tubulin subunits and microtubuleassociated proteins in the central-pair microtubule apparatus of squid (Loligo peallei ) sperm flagella. J. Cell Biol. 89, 309–322. Linck, R. W., Amos, L. A. & Amos, W. B. (1985). Localization of tektin filaments in microtubules of sea urchin sperm flagella by immunoelectron microscopy. J. Cell Biol. 100, 126–135. Linck, R. W., Goggin, M. J., Norrander, J. M. & Steffen, W. (1987). Characterization of antibodies as probes for structural and biochemical studies of tektins from ciliary and 889flagellar microtubules. J. Cell Sci. 88, 453–466. Lupas, A. Van Dyke, M. & Stock, J. (1991). Predicting coiled coils from protein sequences. Science, 252, 1162–1164. Mastronarde, D. N., O’Toole, E. T., McDonald, K. L., McIntosh, J. R. & Porter, M. E. (1992). Arrangement ofinner dynein arms in wild-type and mutant flagella of Chlamydomonas. J. Cell Biol. 118, 1145–1162. McLachlan, A. D. (1982). Rapid comparison of protein structures. Acta Crystallog. 38, 871–873. McLachlan, A. D. & Stewart, M. (1982). Periodic charge distribution in the intermediate filament proteins desmin and vimentin. J. Mol. Biol. 162, 693–695. Meza, I., Huang, B. & Bryan, J. (1972). Chemical heterogeneity of protofilaments forming the outer doublets from sea urchin flagella. Exp. Cell Res. 74, 535–540. Nojima, D., Linck, R. W. & Egelman, E. H. (1995). At least one of the protofilaments in flagellar microtubules is not composed of tubulin. Curr. Biol. 5, 158–167. Norrander, J. M., Amos, L. A. & Linck, R. W. (1992). Primary structure of tektin A1: comparison with intermediate-filament proteins and a model for its association with tubulin. Proc. Natl Acad. Sci. USA, 89, 8567–8571. Norrander, J. M., Linck, R. W. & Stephens, R. E. (1995). Transcriptional control of tektin A mRNA correlates with cilia development and length determination during sea urchin embryogenesis. Development, 121, 1615–1623. Pirner (1995). Molecular associations of tektins. Ph.D. thesis, University of Minnesota, Minneapolis, Minnesota. Pirner, M. A. & R. W. Linck (1994). Tektins are heterodimeric polymers in flagellar microtubules with axial periodicities matching the tubulin lattice. J. Biol. Chem. 269, 31800–31806. Robson, B. & E. Suzuki (1976). Conformational properties of amino acid residues in globular proteins. J. Mol. Biol. 107, 327–356. Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989). Molecular Cloning: A Laboratory Manual, 2nd edit. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Sanger, F., Nicklen, S. & Coulson, A. R. (1977). DNA sequencing with chain–terminating inhibitors. Proc. Natl Acad. Sci. USA, 74, 5563–5567. Staden, R. (1982). An interactive graphics program for

Tektin Structure

comparing and aligning nucleic-acid and amino-acid sequences. Nucl. Acids Res. 10, 2951–2961. Staden, R. (1988). Methods to define and locate patterns of motifs in sequences. Comput. Appl. Biosci. 4, 53–60. Steffen, W. & Linck, R. W. (1988). Evidence for tektins in centrioles and axonemal microtubules. Proc. Natl Acad. Sci. USA, 85, 2643–2647. Steffen, W. & Linck, R. W. (1989a). Tektins in ciliary and flagellar microtubules and their association with other cytoskeletal systems. In Cell Movement (McIntosh, J. R. &. Warner, F. D., eds), vol. 2, pp. 67–81, Alan R. Liss, Inc., New York. Steffen, W. & Linck, R. W. (1989b). Relationship between tektins and intermediate filament proteins: an immunological study. Cell Motil. Cytoskel. 14, 359–371. Steffen, W., Fajer, E. A. & Linck, R. W. (1994). Centrosomal components immunologically related to tektins from ciliary and flagellar microtubules. J. Cell Sci. 107, 2095–2105. Stephens, R. E. (1977). Differential protein synthesis and utilization during cilia formation in sea urchin embryos. Dev. Biol. 61, 311–329. Stephens, R. E. (1989). Quantal tektin synthesis and ciliary length in sea-urchin embryos. J. Cell Biol. 92, 403–413. Stephens, R. E., Oleszko-Szuts, S. & Linck, R. W. (1989). Retention of ciliary ninefold structure after removal of microtubules. J. Cell Sci. 92, 391–402.

397 Stewart, M. (1993). Intermediate filament structure and assembly. Curr. Opin. Cell Biol. 5, 3–11. Stewart, M. & McLachlan, A. D. (1975). Tropomyosin coiled-coil interactions: evidence for an unstaggered structure. J. Mol. Biol. 98, 293–304. Thomas, P. S. (1980). Hybridization of denatured RNA and a small DNA fragments transferred to nitrocellulose. Proc. Natl Acad. Sci. USA, 77, 5201–5205. Weber, K., Plessmann, U., Dodemont, H. & KossmagkStephan, K. (1988). EMBO J. 7, 2995–3001. Weber, K., Plessmann, U. & Ulrich, W. (1989). Cytoplasmic intermediate filament proteins of invertebrates are closer to nuclear lamins than are vertebrate intermediate filament proteins; sequence characterization of two muscle proteins of a nematode. EMBO J. 8, 3221–3227. Witman, G. B., Carlson, K., Berliner, J. & Rosenbaum, J. L. (1972). Chlamydomonas flagella. I. Isolation and electrophoretic analysis of microtubules, matrix, membranes, and mastigonemes. J. Cell Biol. 54, 507–539. Yuan, L., Liu, J.-G. & Ho¨o¨g, C. (1995). Rapid cDNA sequencing in combination with RNA expression studies in mice identifies a large number of male germ cell-specific sequence tags. Biol. Reprod. 52, 131–138.

Edited by M. F. Moody (Received 24 October 1995; accepted 22 December 1995)