Immunological properties and cDNA sequence analysis of an intermediate-filament-like protein from squid neuronal tissue

Intermediate filament proteins are expressed co-ordinately during development and cell differentiation. They are probably involved in special functions related to the differenti-ated state of cells such as the maintenance of resistance to mechanical stress. Some cells lack intermediate filaments, so they do not perform essential ‘house keeping’ functions (Hedberg and Chen, 1986; Bartnik and Weber, 1989).

Intermediate filaments in the cytoplasm of vertebrate cells are composed of different but related polypeptides. Five subgroups for these proteins have been distinguished: type I and II keratins; type III proteins, i.e. vimentin, desmin, GFAP and peripherin; the neurofilament proteins (NFL, NFM, NFH and α-internexin), which constitute the type IV proteins; and the nuclear lamins, which form the type V. All these filamentous proteins share basic structural principles, which are a central α-helical rod region that is flanked by N- and C-domains that are essentially non-helical (Steinert and Roop, 1988). The rod regions form coiled coils-the structural building blocks of the filament back-bone. The nuclear lamins follow similar principles for their construction and are thought to be related through evolution to the intermediate filament family of proteins (Weber et al., 1989; Döring and Stick, 1990).

Sequence information obtained from the intermediate fil-ament proteins of invertebrates shows that they are related more closely to nuclear lamins than to their vertebrate coun-terparts. For example the nuclear lamins of both vertebrates and invertebrates have six extra heptads at the ends of coil 1b and a similar extension to this region has been found in epithelial intermediate filaments from Helix pomatia (Weber et al., 1988), and cytoplasmic intermediate filaments from Ascaris lumbricoides (Weber et al., 1989). Lamin-like tail domains are also found in these invertebrate intermediate filament proteins.

The pattern of similarities is shared by neuronal inter-mediate filaments as well. The two low molecular mass neu-rofilament proteins (60 kDa and 70 kDa) from the squid, Loligo pealei, which have had their sequences deduced from the cDNA sequences, show the basic features of the group of invertebrate intermediate filament proteins (Szaro et al., 1991). The squid neurofilament proteins, like all the vertebrate neurofilaments (Eagles et al., 1990), have long C-terminal extensions to their polypeptides, and this region has clusters of charged residues.

We describe in this paper a new neuronal protein, the cDNA sequence of which was obtained from a squid optic lobe cDNA library. The protein is related to the interme-diate filament class of proteins, yet it is quite distinct from the squid neurofilament sequences so far examined. We have called this protein SNLK (squid neurofilament-like).

The Amersham (UK) cDNA synthesis and cloning system kit (λgt11) was used for the construction of the library. Total RNA was isolated from squid (Loligo forbesi) optic lobes using guani-dinium isothiocyanate as described by Kaplan et al. (1979). Poly(A)⁺ RNA was obtained from the total RNA by oligo(dT) cel-lulose chromatography (Aviv and Leder, 1972). Double-stranded cDNA with EcoRI linkers was prepared by the procedure of Gubler and Hoffman (1983) and inserted into the unique EcoRI site of λgt11 (Huynh et al., 1985); it was then packaged. The total packaged library, before amplification, contained 2.8×10⁸ pfu/ml. Plaques were screened for the presence of neurofilament epitopes essentially as described by Young and Davis (1983). Plaque lifts on nitrocellulose were screened with antibodies generated against purified squid neurofilaments that had been previously dephos-phorylated.

The 3.5 kb cDNA was subcloned into pUC19, and a set of uni-directional nested deletions was generated using exonuclease III then mung bean nuclease (Stratagene, UK) as described by Henikoff (1984). Double-stranded sequencing was carried out using the dideoxy chain-termination method (Sanger et al., 1977). The sequence was analysed with computer programs implemented within the University of Wisconsin Genetics Computer Group (GCG) software on the VAX (Devereaux et al., 1984).

The clone harbouring the 3.5 kb cDNA was used to infect the lysogenic strain E. coli Y1089 as described by Huynh et al. (1985). Fusion proteins were produced by induction with 10 mM IPTG. Bacterial pellets were boiled in sample buffer (1% SDS, 100 mM Tris-HC1, pH 6.8) and spun briefly. The proteins were resolved by SDS-PAGE on a 5-15% gel (Laemmli, 1970) then either stained with Coomassie Blue or electrophoretically transferred onto nitrocellulose (Towbin et al., 1979).

Nitrocellulose filters containing proteins were blocked for 1 hour with 2% (w/w) non-fat milk with TBS (150 mM NaC1, 200 mM Tris-HCl, pH 7) then incubated with primary antibody for 1 hour at room temperature. After three washes with TBS/0.2% Tween-20, peroxidase-conjugated donkey anti-rabbit antibodies were added for polyclonal antibodies or antimouse for monoclonal antibodies (Amersham, UK) diluted 1:1000 in 2% non-fat milk and further incubations carried out for 1 hour. The filters were then washed three times with TBS/0.2% Tween-20 and finally developed with a solution containing 3-amino-9-ethylcarbazole and hydrogen peroxide (Kaplow, 1974).

Anti-SNLK serum and polyclonal antibodies to dephosphorylated squid NFL and NFH were raised in Dutch rabbits by subcutaneous injections of purified antigen. Squid neurofilaments were prepared as described previously (Brown and Eagles, 1986) and dephos-phorylated enzymatically (Chin et al., 1989). Filaments contain-ing all neurofilament-associated polypeptides were used directly for injection without further purification. The antibodies were tested by western blotting and they recognised squid NFH polypeptides and a number of proteins in the region around a rel-ative molecular mass of 60 kDa. The polyclonal serum was freed of any E. coli reacting antibodies by immunoprecipitation with E. coli lysates. For the preparation of anti-SNLK, source gels of the fusion protein β-gal-SNLK were run and the region of the gel con-taining the protein removed. The gel was macerated in adjuvant and then injected. Antigens were injected together with Freund’s complete adjuvant followed by one or two booster injections in incomplete adjuvant. β-galactosidase-reacting antibodies were removed from the serum by immunoprecipitation. IFA (Pruss et al., 1981) was obtained from the ascites fluid of cell lines express-ing this antibody.

Genomic DNA from squid optic lobes was prepared as described by Jeffreys et al., 1977. DNA (5 μg) was digested with restric-tion enzymes then electrophoresed on a 0.8% agarose gel at 90 V. Lambda DNA cut with HindIII was used as markers. Elec-trophoresis of RNA (10 μg) was carried out on a 0.5% agarose/formaldehyde gel. Transfer of DNA and RNA onto nitro-cellulose membranes was as described (Southern, 1975; Thomas, 1980). Filters were prehybridized at 42°C for 3 hours in 50% deionised formamide, 5× SSC, 0.2% SDS, 5 × Denhardt’s solution and 100 μg/ml denatured salmon sperm DNA. For hybridizations, the 3.5 kb cDNA was labelled with [³²P]dCTP (Amersham, UK) by the oligonucleotide-priming procedure of Feinberg and Vogel-stein (1984). Hybridizations were carried out overnight at 42°C. Filters were washed at a final stringency of 0.1× SSC/0.1% SDS at 65°C. Amersham Hyperfilm was used for autoradiography.

Giant axons were dissected out intact from fresh squid, embed-ded in paraffin wax then snap-frozen in liquid nitrogen. Sections (10 μm thick) were cut on a cryostat maintained at a temperature of −20°C. Air-dried sections were fixed with 95% ethanol/5% acetic acid for 10 minutes at −10°C. Sections were blocked in 2% non-fat milk in TBS for 1 hour, washed several times in TBS and exposed for 1 hour at room temperature to anti-SNLK (1:1000 dilution). Sections were then washed with TBS and further incu-bated for 1 hour with rhodamine-labelled goat anti-rabbit anti-bodies (Amersham, UK) diluted 1:50 in blocking solution. After further washes, the sections were layered with an antiphoto-bleaching agent (ascorbic acid, pH 7.0, 5 mg/ml), covered with a coverslip, then viewed using fluorescence microscopy and pho-tographed.

The screening of the λgt11 expression library with squid neurofilament antibodies identified nine positive clones. Of these, three were purified by two rounds of rescreening, and on analysis their insert sizes ranged between 3.0 and 3.5 kb. The clone containing the largest insert was taken for further investigation.

The β-galactosidase fusion protein specified by the 3.5 kb cDNA was generated in amounts sufficient for analysis by SDS-PAGE. Crude protein lysates from the wild-type lysogen clearly showed a protein of around 114 kDa. This protein was identified as β-galactosidase by its binding of an anti-β-galactosidase monoclonal antibody. In the lysate from the recombinant lysogen, the 114 kDa β-galactosidase protein was replaced by a recombinant protein of approximately 180 kDa, indicating that about 66 kDa of polypep-tide is specified by the cloned cDNA. Analysis by Southern blotting (Fig. 1) indicated that the squid genome contained a single copy of the gene. Analysis by northern blotting of squid optic lobe mRNA (Fig. 2) with the 3.5 kb cDNA as probe revealed one major transcript, which had a size of around 2.6 kb.

Crude lysates from the recombinant lysogen were subjected to western blot analysis. The antibodies used were of two types: IFA, a monoclonal antibody that recognises a highly conserved epitope at the end of the rod region of all inter-mediate filaments, and anti-SNF-L, a polyclonal antibody that was raised against the fusion protein specified by the 3.5 kb cDNA. The results are shown in Fig. 3.

The polyclonal serum, after having had all the β-galactosidase antibodies absorbed out, recognizes the fusion protein (Fig. 3, lane d) and the preimmune serum is negative (lane b). The fusion protein is also recognised by IFA (lane c), which indicates that it is related to the intermediate fil-ament class of proteins.

To facilitate sequencing and analysis of the cloned cDNA, the 3.5 kb EcoRI fragment was subcloned into pUC19. Eleven unidirectional deletion mutants were generated by exonuclease III digestion to allow sequencing from vector primers. The complete nucleotide sequence of the cDNA together with the predicted amino acid composition of the protein is shown in Fig. 4. The nucleotide sequence reveals an uninterrupted reading frame covering 557 amino acids. The open reading frame starts with the ATG codon at nucleotides 37 to 39. This ATG is located in an environment propitious for translation because it is located within a sequence differing in only three out of nine positions from the consensus sequence CC(G/A)CCATGG, considered to be a favourable context for eukaryotic translation (Kozak, 1986).

Plots of the α-helical content for the protein and the probability of forming heptad repeats indicated a potential rod region covering residues 88-424. The α-helical pattern search (McLachlan, 1977; Beavil et al., 1992) was also employed to find periodicities within the sequence. This method of analysis is particularly suited to patterns that repeat at regular intervals and which are imperfectly repeated, such as intermediate filaments. Three α-helical regions were defined by the analysis as being significant in terms of coiled-coil-forming ability (Table 1) and they occupied positions consistent with their being coils 1A, 1B and 2 of the rod domain. Of the residues occupying positions ‘a’ and ‘d’ in the rod domain, 63% are hydrophobic. Comparison of the predicted protein sequence of SNLK with other intermediate filament proteins showed that SNLK has many features in common (Table 2 and Fig. 5). A series of heptad repeats covering 337 amino acids define the rod region, leaving a head of 87 amino acids (residues 1 to 87) and a tail extension of 133 amino acids (residues 425 to 557). The rod can be subdivided into smaller sections: coil 1A, 6 heptads; coil 1B, 14 heptads; and coil 2, 21 heptads. Discontinuities in the heptad repeats fall in regions where discontinuities are found in other intermedi-ate filament sequences.

A comparison of the lengths of the various rod domains found in SNLK and other intermediate filaments is shown in Table 2. Coil 1A in SNLK is longer than that seen in SNF60, desmin, intermediate filament from Ascaris, and vertebrate NFL. Coil 1B of SNLK is longer than desmin and NFL but shorter than the lamin-like domains found in the invertebrate intermediate filaments. The length of coil 2 is similar to that found in both the vertebrate and inver-tebrate low molecular mass neurofilament polypeptides and to invertebrate intermediate filament polypeptides.

Most notably the rod region of SNLK has part of the consensus sequence LNDRFANY at the N-terminal end of coil 1A (LNFNNQVI), and at the C-terminal end of coil 2 is found the sequence SRRLLEGSG similar to the IFA epi-tope now identified with the consensus motif YRKL-LEGEE. In summary the predicted protein sequence of SNLK shows the molecule to have all the hallmarks of a typical intermediate filament polypeptide.

To characterise further the protein SNLK, its distribution in various tissues was investigated by using the polyclonal serum raised against it. Western blot analysis was per-formed on squid axoplasm, optic lobe, heart and skin (Fig. 6). The polyclonal serum bound to a single protein of mass around 60 kDa in axoplasm and optic lobe, and no reac-tion was observed with proteins derived from the other tissues. The 60 kDa protein in the neuronally derived tissues also reacted with IFA. SNLK is therefore expressed in cells having a neuronal origin.

The polyclonal serum was also used to stain tissue sec-tions cut from the giant axon (Fig. 7). Thread-like staining within the axoplasm is seen, which indicates that SNLK is probably distributed as filaments in axoplasm.

In order to probe the distribution of molecules resembling SNLK in other species, the antiserum to SNLK was used on western blots containing axoplasm from squid, axoplasm from the marine worm Myxicola infundibulum, partially purified preparations of vertebrate neurofilaments from ox, homogenates from the optic nerve and brain of fish (Oreochromis) and from the clawed toad Xenopus (Fig. 8).

The antiserum reacted with a polypeptide in the squid sample that had an apparent molecular mass of 60 kDa, thus identifying the SNLK molecule. The antiserum also cross-reacted with two polypeptides with molecular masses of around 160 kDa and 170 kDa, corresponding to Myxicola neurofilament polypeptides (Fig. 8, lane B, panel 1). No reaction was observed with mammalian neurofilaments. In addition the antiserum cross-reacted with polypeptides from fish optic nerve, one of which had a molecular mass around 60 kDa.

It therefore seems that SNLK shares immunological similarity with Myxicola neurofilaments but not with mam-malian neurofilaments.

We have cloned the cDNA of a new gene from squid neuronal tissue. The predicted amino acid sequence from the cDNA indicates that the protein is like an intermediate fil-ament polypeptide. On northern blots the major transcript had a size of around 2.6 kb.

Although the serum that we used to screen the library was raised against dephosphorylated squid neurofilaments, it presumably contained antibodies to SNLK as well. Because SNLK and SNF60 have similar molecular masses, the polyclonal serum, when used for western blotting, is not able to distinguish between them and only a single band is seen. Staining with IFA of preparations containing SNLK and SNF60 also shows only one band on blots at the position corresponding to a relative molecular mass of 60 kDa. The relationship between SNLK and the other squid neu-rofilament proteins NF200, NF70 and NF60 is unclear at present. A sequence comparison using the data from Szaro et al. (1991) shows that there is some homology, but there is clearly more similarity between NF200, NF70, and NF60 than between any member of this group and SNLK. SNLK therefore may belong to a separate family. This situation, where a neuron-specific intermediate filament protein is expressed together with a type IV neurofilament protein, is not new and examples include peripherin, preferentially found in neurones of the PNS (Thompson and Ziff, 1989), and XNIF, a developmentally expressed protein in neurones of Xenopus (Charnas et al., 1992). We do not know as yet whether there is a change in the expression of SNLK during development, but we are investigating this possibilty.

Comparison of the predicted amino acid sequence of SNLK with that for other IF proteins demonstrates that SNLK is a true IF protein, with the characteristic head, rod and tail domains, well established for this class of protein. The N-terminal head domain of SNLK is non-α-helical.

This is similar to the situation in invertebrate and vertebrate IF proteins and shows the previously recognised sequence hypervariability (Geisler and Weber, 1982, 1986). This domain also contains a feature that may be important for filament assembly-the sequence RCS at position 58-60.

The motif RXS (where X can be any other residue) is recog-nised by protein kinase A and is found in the same domain of several IF proteins. Mouse NFL for example contains the sequence RTS at position 53-55 and hamster desmin the sequence RTS at position 59-61. With desmin and vimentin, phosphorylation at this head domain site by pro-tein kinase A or protein kinase C blocks subunit polymer-ization in vitro and disassembles pre-existing filaments (Geisler and Weber, 1988; Inagaki et al., 1987). In hamster desmin, mouse NFL, and SNLK, the serine residue is imme-diately preceded by argininesgroups known to be impor-tant in filament assembly (Traub and Vorgias, 1984).

Further evidence in support of the role of the head domain of NFL for filament growth is provided by Gill et al. (1990) who concluded that part of the head domain (between residues 31 and 87) is required for the early steps in assembly. It therefore seems likely that the function served by this site may be conserved in intermediate fila-ments.

In SNLK, charged residues are occasionally found interspersed with hydrophobic residues at postions ‘a’ and ‘d’ within the heptads of SNLK. This, however, is not unique to SNLK because it has been found that periodic bands of alternating positive and negative charge with a period close to 28/3 appear as significant features in the sequences of alpha-keratins (Parry et al., 1977), myosin rod (McLachlan and Karn, 1982) and other intermediate filament proteins. All these proteins form regular assemblies in vivo in which coiled-coils pack side by side. Since there is no extensive sequence homology between these molecules, the 28/3-residue periodicity may reflect a similar pattern of three-dimensional packing.

The percentage homology between each subdomain within the rod region of SNLK and the corresponding domain of the other IF proteins was analysed. Coil 1a had the highest overall sequence homology of all the proteins analysed. The best score of 47% was with the IF B protein from Ascaris lumbricoides. SNLK has two extra heptads in this domain compared to the nuclear lamins and inverte-brate intermediate filaments (Fisher et al., 1986; McKoen et al., 1986; Doring and Stick, 1990; Weber et al., 1989; Szaro et al., 1991). The length of this domain is however similar to the corresponding domain in hamster vimentin, peripherin and α-internexin (Quax et al., 1983; Leonard et al., 1988; Fliegner et al., 1990).

Coil 1b had the lowest overall homology with the other IF proteins. The highest score of 26% is found in mouse NFL and squid (L. pealei) NF60. This domain is peculiar in that nuclear lamins and invertebrate IF proteins possess an extra six heptads (Weber et al., 1989; Doring and Stick, 1990; Fisher et al., 1986; McKeon et al., 1986; Krohne et al., 1987; Gruenbaum et al., 1988; Szaro et al., 1991) when compared to similar domains in vertebrate IFs. SNLK has one heptad more than vertebrate IFs and the recently iden-tified Branchiostoma IF-1 protein (Riemer et al., 1992) and four less than the nuclear lamins and invertebrate cyto-plasmic IFs. The abnormal length of the coil 1b in SNLK suggests that SNLK is a novel neurofilament-like protein from an invertebrate.

Coil 2 of SNLK has 21 heptads. This is in accord with all IF polypeptides sequenced to date. The percentage homology is fairly low but higher than that found for coil 1b. The heptad repeat changes phase at aspartate 303 and glycine 367. Similar positions are found for this phase change in coil 2 for all IF proteins (Weber et al., 1989).

A characteristic feature of the low molecular mass neurofilament polypeptides in vertebrates is a glutamic acid-rich, carboxy-terminal domain (Lewis and Cowan, 1986). These negatively charged residues may be involved in inter-actions with other cytoskeletal proteins within the neurone. These residues are also useful diagnostically because Bodian’s silver stain, a specific marker for neurofilaments, binds to this region (Philips et al., 1983). A group of charged residues is also found in the carboxy-terminal domain of SNLK at positions 498-501. The group consists of three aspartic and two glutamic residues (Fig. 4). This cluster of negatively charged groups may perform similar functions to that found in vertebrate NFL.

In the squid, SNLK is expressed exclusively in tissues having a neuronal origin and here the protein has a fila-mentous distribution within the axoplasm. This distribution is not dissimilar to that observed for neurofilaments, though whether SNLK is attached to the neurofilament network is currently unknown. We searched for the presence of immunologically similar proteins to SNLK in other species. Whilst cross-reactivity was found with neurofilaments from Myxicola, Xenopus and Oreochromis, indicating shared epi-topes with SNLK, no reactivity was found with mammalian neurofilament polypeptides. The epitopes, therefore, have a selected distribution amongst neurofilament species.

To conclude, we have cloned and sequenced the cDNA of a gene for a new protein, SNLK, from squid neuronal tissue. The deduced amino acid sequence of the protein shows that, whilst it shares all the features commonly regarded as being hallmarks of intermediate filaments, it is quite distinct from the neurofilaments sequences examined so far, not only from squid but also from higher species. Further work on SNLK is continuing in order to understand the function of this protein and the structure of the gene.

We thank Mrs R. Rao for providing technical assistance. We also thank Pankaj Marya, Paul Fraylich, Alison Maggs, Bob Hannon, Hitesh Patel, Andrew Beavil and Huub Dodemont for help and discussions throughout this study. We are grateful for the MBA Plymouth for supplying squid and for facilities for car-rying out the initial stages of this study. We also thank the refer-ees for pointing out one of the polyadenylation signals previously missed by us. This work was supported by the SERC and the MRC. The sequence data are available from EMBL/GenBank/DDBJ under the accession number X66695.