Chicken filensin: a lens fiber cell protein that exhibits sequence similarity to intermediate filament proteins

The eye lens is composed of highly specialized cells that focus and transmit light. The lens consists of two principal cell populations, the anterior epithelial cells and the terminally differentiated, posterior fiber cells (for a review see Piatigorsky, 1981). The bulk of the mature lens is composed of elongated, tightly packed fiber cells, which are devoid of most intracellular organelles including the cell nucleus. The cytoplasm of lens fiber cells contains primarily crystallins, a group of water-soluble proteins expressed in very large quantities.

The lens grows throughout the life of the animal by the continual development of new fiber cells around the lens perimeter. Differentiating lens fiber cells express several characteristic proteins, among them the membrane proteins MIP (main intrinsic protein) (Benedetti et al., 1976) and MP70 (Kistler et al., 1985), and the cytoplasmic crystallins (Piatigorsky, 1981). Elongating cells are gradually displaced toward the lens center, concurrent with the loss of cytoplasmic organelles. The mature cells of the central fiber cell mass, devoid of nuclei and protein synthesizing machinery, persist for the life of the animal.

The lens cell cytoskeleton is made up of typical microtubules, microfilaments and vimentin intermediate filaments (Ramaekers et al., 1980). Recent reports suggest that cytokeratins may also be present in the epithelium-derived lens (Magin et al., 1990; Bader, B. L., Magin, T. M. and Franke, W. W. (1992) Abstract: Fifth International Congress on Cell Biology, p. 79). Intermediate filaments are built from a diverse family of proteins, the common element of which is a central rod domain, conserved in secondary structure and to a lesser extent in amino acid sequence. The intermediate filament rod consists of 310 amino acids (356 for nuclear lamins and invertebrate cytoplasmic intermediate filament proteins) dominated by a heptad repeat with characteristically spaced interruptions in the heptad pattern (Geisler and Weber, 1982; Conway and Parry, 1988; Steinert and Roop, 1988). The rod is divided into three α-helical segments, coils 1A (35 amino acids), 1B (101 amino acids) and 2 (148 amino acids) separated by two non-helical linkers of variable lengths. A stutter in the heptad repeat phasing marks the middle of coil 2. On the basis of sequence similarities, intermediate filament proteins have been classified into six major types (Conway and Parry, 1988; Lendahl et al., 1990). Rod domain amino acid sequence homology between members of an intermediate filament protein type ranges from 70 to 95% identity, whereas rod domain amino acid sequence identity is 20 to 30% between members of different intermediate filament protein types (Steinert and Roop, 1988).

Lens fiber cells express two antigenically unique, cytoskeletal proteins, CP95 and CP49 in chicken (Ireland and Maisel, 1984). These lens-specific proteins have been immunolocalized to beaded filament structures observed by electron microscopy (Ireland and Maisel, 1984, 1989). This paper reports the cloning and sequencing of the full-length cDNA for chicken filensin, or CP95, and demonstrates the sequence similarities between filensin and the intermediate filament family of proteins.

The lenses of 13-day embryonic White Leghorn/Rhode Island Red hybrid chickens were dissected and frozen immediately in liquid nitrogen. Total RNA was extracted from the lenses using a modification of a guanidinium thiocyanate protocol (Chirgwin et al., 1979). Poly(A)⁺ RNA was isolated over an oligo(dT)-cellulose (Collaborative Research) column (Maniatis et al., 1982).

A cDNA expression library was constructed in λgt11 from 13-day embryonic chicken lens poly(A)⁺ RNA using a modification of the protocol described by Huynh et al. (1985). First- and second-strand syntheses were performed as described by Watson and Jackson (1985). The Protoclone™ λgt11 System (Promega) was used for ligation of cDNAs to λgt11 arms and for phage packaging. The library was amplified once prior to screening.

Polyclonal antibodies, RQR and 6182, were used to screen the β-galactosidase fusion proteins expressed by the λgt11 library clones (Snyder and Davis, 1985). Modifications to the protocol were the following: the blocking agent was 1% dried milk powder and the secondary antibody was horseradish peroxidase-conjugated goat anti-rabbit IgG (BMB). Peroxidase was detected colorimetrically (Young and Davis, 1985).

RQR (Menko et al., 1984) and 6182 were rabbit polyclonal antibodies raised against chicken lens MIP (MP28). The immunogen was prepared by separating lens membrane proteins in SDS-PAGE and cutting the 28 kDa MIP band from the gel. RQR and 6182 contained contaminating reactivity to chicken filensin.

Affinity-purified anti-filensin antibody was obtained by eluting the polyclonal antibody 6182 from protein expressed by λgt11 clone C5 (Snyder and Davis, 1985). Immunoblots of chicken lens proteins indicated that the resultant affinity-purified antibody was specific for filensin.

MA₂ (kindly provided by H. Maisel) was a mouse monoclonal antibody raised against chicken CP95. MA₂ recognized the protein expressed by chicken lens λgt11 clone C5 (filensin).

#15, a mouse monoclonal antibody raised against bovine lens MIP, recognized chicken lens MIP.

λgt11 DNA was isolated by PEG precipitation (Maniatis et al., 1982) and digested with EcoRI. Inserts were subcloned into the EcoRI site of pBluescript II KS^- (Stratagene). Deletion subclones were generated using exonuclease III (New England Biolabs) (Henikoff, 1987). Double-stranded plasmid DNA was sequenced using the Sequenase^® Version 2.0 DNA Sequencing Kit (United States Biochemical) with T3 and T7 promoter-specific primers and [α-³⁵S]dATP (Amersham).

FastDB (Brutlag et al., 1990), an IntelliGenetics, Inc. sequence analysis computer program, was used to search all entries in the GenBank and EMBL nucleotide sequence data bases and in the PIR and SwissProt amino acid sequence data bases for sequence homology to C5. Amino acid sequence comparisons were conducted using a PAM 256 mutation probability matrix (Dayhoff et al., 1972) and a structure-genetic matrix (McLachlan, 1972). PILEUP (Devereux et al., 1984), a GCG sequence analysis software program (version 7.1), was used to align the deduced amino acid sequence of chicken filensin with the rod domain sequences of 15 intermediate filament proteins. The gap open parameter was 4.0; the gap elongation parameter was 0.1. A subset of this alignment is presented in Fig. 2 (below).

RNA samples were denatured in 50% formamide and 6% formaldehyde in low ionic strength buffer (20 mM phosphate, pH 7.7) at 60°C for 5 min. The samples were electrophoresed through a 1% agarose gel containing 6% formaldehyde in 20 mM phosphate buffer, pH 7.7. Ribosomal bands were visualized by staining a portion of the gel with 0.5 µg/ml ethidium bromide. The RNA in the remainder of the gel was capillary transferred to nitrocellulose from 10× SSC (1.5 M sodium chloride, 0.15 M sodium citrate, pH 7.0). Prehybridization and hybridization were in 50% formamide at 42°C (Maniatis et al., 1982). The filter was washed several times in 1× SSC, 0.1% SDS and finally in 0.2× SSC, 0.1% SDS at 60°C.

Probe: pUC119 containing a 2.0 kb EcoRI fragment of C5, bases 84 through 2114, was nick translated in the presence of [α-³²P]dCTP (Amersham) (Maniatis et al., 1982). Incorporated nucleotides were separated from unincorporated nucleotides in a BioGel P-60 (Bio-Rad) column. The labeled probe was denatured by boiling 5 min in a water bath and added to fresh hybridization solution at 1×10⁶ c.p.m./ml.

RACE (rapid amplification of cDNA ends) using PCR (polymerase chain reaction) (Frohman, 1990) was performed. Reverse transcription of poly(A)⁺ embryonic chicken lens RNA with AMV reverse transcriptase (Promega) was primed with a synthetic oligonucleotide primer complementary to bases 498 to 517 of the chicken filensin cDNA sequence (5’-TTCGTTGGCG-GTTGAACTCG-3’) (National Biosciences). Primers for the PCR were as follows: 25 picomoles adapter primer (5’-GACTCGAG-GATCCAAGC-3’), 10 picomoles oligo(dT)₁₇-adapter primer (gifts from S. Gantt, University of Minnesota, USA), and 25 picomoles of sequence-specific synthetic oligonucleotide primer complementary to bases 372 to 391 of chicken filensin and flanked by an EcoRI restriction enzyme site (5’-CTGAGTGAATTCG-GTTGATGTAGCTGGCGAAG-3’) (National Biosciences). The reaction mixture was denatured at 94°C for 5 min and cycled 30 times in a DNA Thermal Cycler (Perkin Elmer) as follows: 94°C for 1 min/50°C for 2 min/72°C for 3 min. The extension step was lengthened by 5 s each cycle, and the last cycle was incubated an additional 10 min at 72°C. cDNA PCR products were digested with BamHI and EcoRI, cloned into BamHI/EcoRI-digested pBluescript II KS^-, and sequenced.

Genomic DNA from brains of 13-to 15-day embryonic White Leghorn chickens (inbred at the University of Minnesota) was isolated and dialyzed against several changes of TE (Maniatis et al., 1982). Samples (10 µg) of genomic DNA were digested overnight in the appropriate buffers with each of the following restriction enzymes; BamHI, HindIII, SacI, XhoI (New England Biolabs), EcoRI (BMB) and XbaI (BRL). The digested DNA was elec-trophoresed through a 0.7% agarose gel, capillary transferred to Zeta-Probe^® (Bio-Rad), and UV crosslinked to the membrane (Stratagene). Prehybridization (30 min) and hybridization (16 h) were performed as per the formamide protocol for Zeta-Probe^® Blotting Membrane (Bio-Rad) at 43°C. The membrane was washed several times in 1× SSC, 0.1% SDS and in a final concentration of 0.2× SSC, 0.1 % SDS at 50°C.

Probe: a 300 base pair PstI fragment spanning bases 53 to 352 of chicken filensin cDNA sequence was radiolabeled with [α-³²P]dCTP (Amersham) using the Random Primers DNA Labeling System (BRL). Incorporated nucleotides were separated from unincorporated nucleotides in a BioGel P-60 (Bio-Rad) column. The labeled probe was boiled for 5 min and added to hybridization solution at 1×10⁷ c.p.m./ml.

Total cell proteins from embryonic and adult chicken lenses and from embryonic brain, intestine, liver, heart and skin were solubilized by sonicating in PBS (2.7 mM KCl, 1.5 mM KH₂PO₄, 137 mM NaCl, 8 mM Na₂HPO₄, pH 7.2) with 0.1% SDS, 0.3 mM PMSF, 1.5 mM EDTA, 0.5 µg/ml leupeptin and 1 µg/ml pepstatin. Protein concentrations were determined with a BCA protein assay (Pierce). Protein samples were boiled for 5 min in loading buffer, separated by SDS-PAGE (Laemmli, 1970), and electrophoretically transferred (Bio-Rad Trans-blot Cell) to BioTrace™ NT membrane (Gelman) in transfer buffer (25 mM Tris-base, 192 mM glycine, 0.01% SDS, 10% methanol). The blots were blocked in 2% BSA in TBS (50 mM Tris-HCl, pH 7.5, 150 mM NaCl) and incubated overnight with affinity-purified anti-filensin antibody 6182. The secondary antibody was biotinylated goat anti-rabbit IgG (Kirkegaard & Perry Laboratories, Inc.) (1:1000) followed by horseradish peroxidase-conjugated streptavidin (Molecular Probes, Inc.) (1:2000). Peroxidase was detected colorimetrically by incubating the filter for 20 min in 1 volume of 0.3% 4-chloro-1-napthol (Sigma) dissolved in cold methanol to which 1.5 volumes of TBS containing 0.05% hydrogen peroxide was added. Molecular weight standards were biotinylated (Bio-Rad) or SDS-PAGE Molecular Weight Protein Standards - High (Bio-Rad).

cDNA clone C5, bases 84 to 2624 of chicken filensin cDNA sequence in the EcoRI site of pBluescript II KS^-, was linearized with XhoI (New England Biolabs). A coding strand transcript was synthesized with 25 units of T7 RNA polymerase (Stratagene) using the Riboprobe^® Gemini System II (Promega) reagents and protocol. In vitro translation of the synthetic RNA transcript of C5 and of embryonic chicken lens poly(A)⁺ RNA was performed in rabbit reticulocyte lysate (Promega) in the presence of [³⁵S]methionine (Amersham). The ³⁵S-labeled proteins were separated with SDS-PAGE in an 8% gel (Laemmli, 1970), the gel was soaked for 30 min in 1 M sodium salicylate, and the labeled protein bands were visualized by exposure to Kodak X-OMAT^AR film at-70°C.

Samples (40 µl) of ³⁵S-labeled, in vitro translation products from chicken lens RNA or from the synthetic C5 transcript were incubated with 20 µl of each of three antibodies (polyclonal antibody RQR, or monoclonal antibodies MA₂ or #15) for 30 min on ice in the presence of 10 mM EDTA and 1 mM PMSF. A 100 µl sample of buffer-washed Protein A-Sepharose CL-4B (Sigma) for RQR, or 100 µl of buffer-washed Protein A plus Protein GSepharose (Sigma) for MA₂ and #15, were added to the antigenantibody complexes in 500 µl buffer (25 mM Tris-HCl, pH 8.0, 100 mM NaCl, 0.5% NP-40, 2 mM EDTA, 1 mM PMSF) and incubated for 1 h at 4°C. The Sepharose beads were washed several times in the same buffer and the antigen-antibody complexes were released from the beads by boiling for 5 min in SDS-PAGE loading buffer (Laemmli, 1970).

Lens ‘half shells’

Whole lenses were dissected from 11- and 14-day embryonic chickens, fixed for 2-3 h in 2.5% paraformaldehyde in PBS, and medially transected from the lens anterior to the posterior (as suggested by W. T. M. Gruijters, University of Auckland, New Zealand). The lightly fixed central cellular masses were discarded leaving lens half shells consisting of the lens capsule, the epithelial cell layer, peripherally located fiber cells, and ends of more centrally located fiber cells abutting on the epithelial cell layer anteriorly and the lens capsule posteriorly.

Whole lenses dissected from 11- and 14-day embryonic chickens were fast fresh frozen in OTC in dry-ice-chilled isopropanol. Lenses were sectioned (8-10 µm), mounted on slides coated with a 2% solution of 3-aminopropyltriethyl oxysilane in acetone and fixed in 2.5% paraformaldehyde for 30 min.

Lens half shells and lens sections were incubated with affinity-purified anti-filensin antibody 6182 overnight at 4°C, then with FITC-conjugated anti-rabbit IgG (1:100) (BMB) overnight at 4°C. Lens shells and sections were mounted in Aquamount and photographed on a Zeiss fluorescence microscope.

I constructed a cDNA expression library in λgt11 using poly(A)⁺ lens RNA isolated from embryonic White Leghorn/Rhode Island Red hybrid chickens. I screened the library with two polyclonal antibodies, RQR and 6182, each raised against the 28 kDa chicken lens MIP. Five independent cDNA clones were initially characterized by probing immunoblots of chicken lens proteins with antibody eluted from protein expressed by each of the five cDNA clones. Antibody eluted from four of the clones recognized a 102109 kDa triplet of chicken lens proteins distinct from MIP (data not shown).

Fig. 1 displays the complete cDNA sequence and the deduced amino acid sequence of the 102-109 kDa chicken lens protein. The nucleotide sequence was determined from C5, a nearly complete λgtll cDNA clone, and from the 5’ extension of C5 with RACE-PCR (Frohman, 1990). Comparison of the C5 nucleotide sequence with the GenBank and EMBL data base entries indicated that C5 was a novel sequence. A preliminary report of the C5 sequence has been presented (Remington, S. G. (1990) J. Cell Biol. 111, 44a).

I sequenced both strands of C5, bases 84 through 2624 (Fig. 1). C5 was flanked at the 3’ end by a poly(A)+ tail at base 2609, which was preceded by the polyadenylation signal AAUAAA at base 2586. The longest open reading frame in C5 extended from the predicted methionine residue at base 229 to the stop codon at base 2200 and was preceded by two in-frame stop codons (bases 100 and 115). The nucleotide sequence immediately surrounding the presumed start of translation, GAAGCCAUGU, conformed reasonably well to the vertebrate translation initiation consensus sequence, GCC(A/G)CCAUGG (Kozak, 1991). The open reading frame predicted a 657 amino acid protein with a molecular mass of 76,110 Da, significantly smaller than the molecular masses estimated from the migration of the embryonic chicken protein in SDS-PAGE, 102-109 kDa (see Fig. 6A below).

C74, another 102-109 kDa cDNA clone identified in the λgt11 library by hybridization to C5, extended from base 110 through a 3’ poly(A)+ tail of 23 bases. The nucleotide sequence of C74 differed from that of C5 at eight isolated base positions. The C5 base number and nucleotide from Fig. 1 and followed by the corresponding C74 nucleotide: 152T♦C, 396G♦A, 489T♦C, 593A♦G, 1828T♦A, 2371C♦T, 2606C ♦A and 2608G♦A. Two of the base differences between C5 and C74 predicted single amino acid changes: H to R at amino acid 122 and F to I at amino acid 534. Clones C5 and C74 may have represented polymorphic variants in the 102-109 kDa protein of the two chicken species, White Leghorn and Rhode Island Red, which were bred to produce the hybrid chickens used for the lens library construction. Southern data from inbred White Leghorn genomic DNA indicated that the 102-109 kDa protein was encoded by a single copy gene (see below).

I compared the predicted amino acid sequence of C5 with the PIR and SwissProt amino acid sequence data base entries using FastDB (Brutlag et al., 1990) (IntelliGenetics, Inc.). Comparison of the amino-terminal half of the C5 translation, allowing conservative amino acid substitutions, revealed weak sequence homology with the intermediate filament family of proteins. Type IIb cytokeratins produced the highest initial alignment scores, whether the data base searches were performed using a mutation probability PAM256 matrix (Dayhoff et al., 1972) or a structure-genetic matrix (McLachlan, 1972). Data base comparisons of the carboxy-terminal half of C5 did not reveal significant sequence homology to nucleotide (GenBank, EMBL) or protein (PIR, SwissProt) data base entries. A partial cDNA sequence encoding the equivalent rat lens protein, CP94, has recently been published (Masaki and Watanabe, 1992).

The sequence data, suggesting that C5 could encode a cytoskeletal protein, prompted comparison of the C5 sequence with that of cytoskeletal bovine filensin (Gounari et al., 1993). The amino-terminal halves of the chicken and bovine deduced amino acid sequences were greater than 60% identical, and the carboxy-terminal 45 amino acids were 82% identical. Predicted amino acids 625 to 642 of chicken C5 were identical at 14 of 18 residues to a car-boxy-terminal bovine filensin proteolytic peptide, bovine filensin amino acids 718 to 749 (Gounari et al., 1993). Sequence similarity of clone C5 with bovine filensin indicated that C5 probably encoded the avian analogue of filensin. I refer to the 102-109 kDa protein encoded by C5 as chicken filensin.

The predicted amino acid sequence of chicken filensin contained a 280 residue domain, amino acids 39 to 318, dominated by a heptad repeat pattern. Fig. 2 illustrates an alignment (Devereux et al., 1984) of the rod domains of chicken filensin and three representative intermediate filament proteins, a mouse type II cytokeratin (k2c1) (Steinert et al., 1985), human glial fibrillary acidic protein (GFAP) (Reeves et al., 1989), and chicken middle-molecular weight neurofilament protein (NF-M) (Zopf et al., 1990). Note the similarity in the positions of hydrophobic amino acids, indicated in bold type. Alignment of the 280 amino acid rod domain of chicken filensin with the 310 amino acid rod domain of intermediate filament proteins was optimized by inserting a gap in the filensin sequence in the amino-terminal half of coil 2.

The amino-terminal 35 residues of the chicken filensin rod domain, amino acids 39 to 73, exhibited 45% amino acid identity with GFAP intermediate filament coil 1A; chicken filensin residues 45 to 63 were 68% identical to the highly conserved internal coil 1A motif of GFAP (Conway and Parry, 1988). Domains 1A and 1B of chicken filensin were separated by a nine amino acid interruption in the heptad repeat pattern, corresponding to intermediate filament linker L1. The next 101 amino acids of filensin were 30% identical to mouse k2c1 domain 1B and were followed by a second interruption in the heptad repeat pattern, coinciding with intermediate filament linker L12.

Alignment of the chicken filensin sequence with intermediate filament segment 2 required a 29 amino acid gap in the filensin sequence (Fig. 2). Chicken filensin residues 201 to 224 were 24% identical to the corresponding sequence of NF-M; filensin residues 225 to 318 were 19% identical to the corresponding sequence of GFAP. The heptad repeat phasing of the putative filensin rod domain shifted at amino acid 256, dropping three residues of a heptad repeat as occurs at the stutter in the middle of intermediate filament coil 2. The filensin sequence maintained the intermediate filament spacing of the stutter in relation to the carboxy-terminal end of the rod. Chicken filensin residues 300 to 318, allowing conservative amino acid substitutions, contained the conserved sequence at the carboxy terminus of the intermediate filament coil 2.

The alignment depicted in Fig. 2 predicts that chicken filensin proline residue 217 lies within α-helical rod segment 2. Proline residues of intermediate filament proteins have not been found within α-helical segments. However, the introduction of proline residues in internal coiled-coil rod segments of human keratin K14 resulted in polymerization competent molecules under the experimental conditions (Letai et al., 1992). The keratin K14 results suggested that proline residues may be tolerated within internal α-helical regions of the intermediate filament rod domain.

The 280 amino acid chicken filensin rod domain was flanked by a 38 residue amino-terminal head domain and by a relatively large, 339 residue carboxy-tail domain. The head and tail domains of chicken filensin contained no significant sequence similarity to other intermediate filament proteins, or to other sequences in the data bases.

The overall identity between the putative rod domain sequence of chicken filensin and that of individual intermediate filament proteins ranged from 20 to 25%. However, filensin contained the smallest intermediate filamentlike rod domain sequenced to date. In summary, the predicted amino acid sequence of chicken filensin contained striking similarities to the rod domain of intermediate filament proteins, suggesting that filensin was related to intermediate filament proteins.

To determine the size of chicken filensin messenger RNA, I probed a blot of total embryonic chicken lens RNA with a C5 EcoRI fragment (bases 84 to 2114) encompassing most of the chicken filensin coding region. The C5 cDNA hybridized to a prominent 2.7 kb band and to a fainter 3.0 kb band (Fig. 3). Poly(A)⁺ embryonic chicken lens RNA, probed with either a 750 base 5’Bam HI restriction fragment (bases 84 to 872) or an EcoRI restriction enzyme fragment encompassing the 3’ untranslated region of chicken filensin (bases 2114 to 2624), yielded the same 2.7 kb prominent band and 3.0 kb minor band (data not shown). The filensin cDNA sequence in Fig. 1 agreed with the 2.7 kb messenger RNA size estimate, assuming a 100 to 200 base poly(A) ⁺ tail on lens RNA molecules.

To obtain the complete 5’ end of the chicken filensin mRNA sequence, I used the RACE-PCR protocol (Frohman, 1990). I primed first-strand cDNA synthesis of chicken lens RNA with a synthetic oligonucleotide complementary to bases 498 to 517 of the filensin sequence. PCR amplification of the cDNA was primed with a filensin sequence-specific oligonucleotide, complementary to bases 372 to 391. The major PCR product was a 400 base pair band, which hybridized to a filensin sequence probe. A minor PCR product at 850 base pairs, not visible with ethidium bromide staining, also hybridized to the filensin probe (data not shown).

I cloned the chicken filensin 5’ extended products and sequenced inserts that hybridized to a 5’ filensin probe. The resultant nucleotide sequence from several different PCR product clones extended the C5 filensin sequence 15 bases 5’ (denoted ‘PCR’ in Fig. 1). This 5’ extension of the chicken filensin sequence was obtained from at least four independent cDNAs. Each of the two polymorphic differences at base 152 was represented in the PCR product clones derived from two different cDNA populations. A single PCR product clone extended the filensin cDNA sequence an additional 69 bases 5’ (to base 1 of Fig. 1). The sequence of the major chicken filensin PCR product, which migrated as a 400 base pair band, suggested that λgt11 clone C5 was nearly full length.

I also sequenced a 2.5 kb White Leghorn chicken genomic subclone that overlapped with the first 586 bases of filensin cDNA sequence and extended an additional 1.8 kb 5’ (data not shown). In the region of genomic and cDNA sequence overlap, the filensin C5 and corresponding PCR product sequences exactly matched the White Leghorn genomic sequence, with the exception of a G (not shown) invariably positioned on the 5’ end of the PCR product cDNAs. Clone C5 probably represented a White Leghorn polymorphic variant of chicken filensin.

To determine the copy number of the filensin gene in the chicken genome, I hybridized a chicken filensin PstI fragment, bases 53 to 352 of the cDNA sequence in Fig. 1, to a blot of restriction enzyme-digested, White Leghorn genomic DNA. The labeled PstI fragment hybridized to a single DNA fragment in each of the genomic digests, indicating that filensin probably was a single copy gene (Fig. 4).

Western immunoblots of proteins from embryonic and adult chicken lenses and from several embryonic chicken organs suggested that filensin was a lens-specific protein. I electrophoresed total cell protein from hybrid embryonic chicken lens, brain, intestine, liver, heart and skin in a 12% SDS-polyacrylamide gel, transferred the proteins to nitrocellulose, and probed the blot with affinity-purified antifilensin antibody 6182 (Fig. 5). Chicken filensin migrated as a band with an apparent molecular mass of approximately 105 kDa in the lens protein lane. Affinity-purified 6182 did not recognize proteins from any other chicken organ examined. Affinity-purified 6182 did not retain significant reactivity to MIP, the lens membrane protein to which the antibody was made (note the absence of reactivity in the 28 kDa region of the lens lane).

In a 7.5% SDS-polyacrylamide gel, I electrophoresed total lens protein from embryonic White Leghorn chickens and from one-year-old White Leghorn roosters, transferred the proteins to nitrocellulose, and probed the blot with affinity-purified anti-filensin antibody 6182 (Fig. 6A). Anti-filensin antibody recognized a triplet of embryonic chicken lens proteins, 102, 105 and 109 kDa, that were not resolved in a 12% gel. In adult rooster lens protein, anti-filensin antibody recognized a doublet with apparent molecular masses of 105 and 109 kDa, as well as a minor broad band at 40 kDa (Fig. 6A). The lens protein samples were prepared on ice in the presence of several protease inhibitors. The different chicken filensin protein profiles in embryonic lens versus rooster lens suggested that filensin was subject to developmentally regulated modifications.

The open reading frame in the cDNA sequence of chicken filensin predicted a protein with a molecular mass of 76 kDa; however, the apparent molecular mass of chicken filensin in SDS-PAGE was 102-109 kDa. The molecular mass discrepancies could have resulted from post-translational modifications of filensin and/or an incomplete cDNA containing a sequencing error that resulted in a frame shift. If cDNA clone C5 contained the entire coding region for chicken filensin, then in vitro transcription/translation of clone C5 should yield the same protein product(s) as in vitro translation of embryonic chicken lens RNA.

Fig. 6B shows a comparison of the clone C5 in vitro transcription/translation product(s) with embryonic chicken lens RNA in vitro translation/anti-filensin immunoprecipitation product(s). Negative and positive controls for the in vitro translation reactions are shown in lanes 8 and 9. In lane 8, a negative control (no added RNA) generated no bands. In lane 9, a positive control (embryonic chicken lens RNA) generated a characteristic chicken lens protein pattern dominated by delta crystallin at 50 kDa (see arrowhead at right in Fig. 6B).

In vitro translation of the C5 transcript (lane 7) demonstrated a protein doublet with apparent molecular masses of 102 and 105 kDa. The observed doublet may have resulted from post-translational modification that was supported in the rabbit reticulocyte lysate (i.e. protein kinase activity) or from translation initiation at two different methionine residues. Multiple translation initiations were an unlikely explanation for the presence of two bands. After the translation initiation methionine at base 229, the next AUG codon was at base 454, a position beyond the intermediate filament rod coil 1A homology. Also, use of the initiation codon at base 454 would predict a molecular mass difference of 9 kDa compared with the observed difference in mobility of 3 kDa.

Identification of the filensin in vitro translated product(s) from chicken lens RNA required immunoprecipitation with anti-filensin antibodies. I performed separate immunoprecipitations (lanes 4-6) from the embryonic chicken lens RNA in vitro translation products (represented in lane 9) using each of the following antibodies: #15, a monoclonal antibody to MIP; MA₂, a monoclonal antibody to CP95 (filensin); and RQR, a polyclonal antibody raised against chicken lens MIP and containing reactivity to filensin. SDS-PAGE and autoradiography revealed a protein doublet at 102 and 105 kDa in samples that were immunoprecipitated with either MA₂ or RQR (lanes 5 and 6). This doublet appeared similar to the doublet generated by in vitro translation of the C5 transcript (lane 7).

A prominent band migrated with the dye front from the in vitro translated products of chicken lens RNA immunoprecipitated with RQR or #15 antibodies (data not shown). Each of the antibodies RQR and #15 was raised against lens MIP, and the dye front bands probably contained chicken lens MIP. #15 antibody did not immunoprecipitate the 102 and 105 kDa protein doublet (lane 4).

To verify that the C5 transcript protein doublet was the same as the immunoprecipitated lens RNA in vitro translated doublet, I performed separate immunoprecipitations (lanes 1-3) from the C5 transcript product(s) (represented in lane 7) using the same three antibodies, #15, MA₂ and RQR. SDS-PAGE and autoradiography revealed the identical doublets in samples that were immunoprecipitated with either MA₂ or RQR (lanes 2 and 3). #15 antibody did not immunoprecipitate any ³⁵S-labeled protein from the C5 transcript products (lane 1). The in vitro translated doublet from the C5 transcript migrated slightly faster with unlabeled rabbit reticulocyte lysate proteins (lane 7) than in the immunoprecipitation lanes with unlabeled antibodies (lanes 2 and 3). The ³⁵S-labeled product was the same in these three lanes; the difference among the lanes was the overall protein content.

Careful alignment of an anti-filensin immunoblotted lane of authentic embryonic lens protein and the autoradiograph of in vitro translated products from the same gel indicated that the in vitro translated doublet represented the lower two bands of the embryonic chicken lens filensin triplet (data not shown). In contrast, the inherent rooster doublet comigrated with the upper two bands of the embryonic triplet (Fig. 6A). The presence of different subsets of 102109 kDa bands in vivo and in vitro suggested that chicken filensin was subject to differential post-translational modification (i.e. covalent amino acid modification(s) and/or protease cleavages) in embryonic chicken lens, in adult rooster lens and, in vitro, in rabbit reticulocyte lysate.

Production of the same in vitro translated protein products from the clone C5 transcript and from chicken lens RNA implied that clone C5 contained the entire filensin coding region. The anomalous migration of chicken filensin in SDS-PAGE (i.e. with apparent molecular mass of 102105 kDa rather than the calculated 76 kDa) may have resulted from the primary amino acid sequence alone and/or from post-translational modifications that occurred in the rabbit reticulocyte lysate and in the lens.

Immunofluorescence was used to localize the filensin antigen in chicken lenses. Whole embryonic chicken lenses were cut in half transversely (anterior to posterior of the lens), which resulted in the extrusion of the central lens core. The remaining half shells were composed of the lens capsule, the anterior epithelial cell layer, elongating fiber cells of the equatorial region and fiber cell ends in more central regions. The lens half shells were incubated with affinity-purified anti-filensin antibody 6182. Fig. 7A illustrates the predominant anti-filensin image of labeled fiber cell ends abutting on the posterior capsule or the anterior lens epithelial cell layer. Labeled membranes of elongated fiber cells were occasionally seen extending from the lens half shell (Fig. 7B).

In frozen sections of embryonic chicken lenses, fiber cell membranes were labeled with affinity-purified anti-filensin antibody 6182 (Fig. 7C and D). Note the absence of label in adjacent lens epithelial cells (*), and also in the equatorial region (Fig. 7C). In many sections the fiber cell membrane labeling appeared more intense in punctate spots along the boundaries that abutted on the posterior capsule and the anterior epithelial cell layer (Fig. 7C). Anti-filensin antibodies labeled lens fiber cell membranes, including fiber cell ends, but did not label lens epithelial cells.

Screening an embryonic chicken lens library, I cloned and sequenced a novel cDNA that encoded filensin, a lensspecific protein with sequence homology to the rod domain of intermediate filament proteins. RNA blot analysis and RACE-PCR indicated that the major lens mRNA for chicken filensin was 2.7 kb. The single longest open reading frame within the chicken cDNA sequence predicted a protein product with a molecular mass of 76 kDa. Authentic lens filensin and in vitro translated filensin migrated in SDS-PAGE as multiple protein bands with apparent molecular masses between 102 kDa and 109 kDa.

In vitro translation of a synthetic transcript of C5, the longest isolated cDNA clone, or in vitro translation of embryonic lens RNA yielded the same 102 and 105 kDa molecular mass protein doublet. Generation of the same in vitro translated products from the C5 transcript and from embryonic chicken lens RNA supported the conclusion that C5 contained the entire coding region for chicken filensin.

Despite the presence of two upstream and in-frame stop codons, λgt11 clone C5 produced a protein product in Escherichia coli with the antigenicity of chicken filensin. The protein could have been a β-galactosidase fusion protein if the termination codons TAA at base 100 and a TGA at base 115 were ignored. Alternatively, the bacteria may have reinitiated translation with the predicted methionine codon at base 229, or with another internal AUG codon in the chicken filensin cDNA sequence.

Antibody reactivity indicated that chicken filensin was cytoskeletal protein CP95 (Ireland and Maisel, 1989). CP95 was one of two urea-soluble, chicken lens fiber cell-specific proteins of 95 kDa and 49 kDa, CP95 and CP49 respectively, that cofractionated with vimentin and actin (Ireland and Maisel, 1984). Monoclonal antibody MA₂ (raised against CP95 and kindly provided by H. Maisel) recognized protein expressed by chicken filensin clone C5, as well as the multiple 102-109 kDa protein bands from embryonic and adult rooster lenses (data not shown). I used MA₂ in this study to immunoprecipitate chicken filensin from the in vitro translated products of chicken lens RNA and the C5 synthetic transcript.

Nucleotide and peptide sequence comparisons suggested that chicken filensin, or CP95, is the chicken equivalent of bovine filensin. Comparison of the chicken and the bovine predicted amino acid sequences (Gounari et al., 1993) indicated greater than 60% identity between the amino-terminal halves of the molecules, encompassing the heptad repeat domains of the filensins, and 82% identity between the car-boxy-terminal 45 amino acids of chicken and bovine filensin. Comparison of the chicken filensin predicted amino acid sequence with the published rat CP94 partial amino acid sequence (Masaki and Watanabe, 1992) indicated similar levels of sequence homology within the heptad repeat domains (approximately 60%), but little amino acid homology in the carboxy-terminal tails. Predicted amino acid sequence homology between the carboxy-terminal tails of chicken filensin and rat CP94 can be found in another reading frame of the rat nucleotide sequence.

Biochemical attributes and antibody crossreactivity support the conclusion that CP95 is chicken filensin. Comparable lens cell fractionation procedures for chicken (Ireland and Maisel, 1989), bovine (FitzGerald and Gottlieb, 1989) and porcine lenses (Merdes et al., 1991) yielded lensspecific urea-soluble proteins with estimated molecular masses of 95 kDa for chicken CP95, 115 kDa for bovine, and 100 kDa for porcine filensin. Antibody crossreactivity implied that the chicken CP95 and bovine 115 kDa were related proteins (FitzGerald, 1988a), and that bovine 115 kDa and porcine filensin were analogous proteins (S. D. Georgatos, personal communication).

Filensin is apparently lens-specific. Anti-filensin antibodies recognized protein bands in immunoblots of chicken lens, but not in immunoblots of other embryonic chicken organs including skin, heart, liver, intestine and brain (Fig. 5). Based on the absence of antibody reactivity with proteins from other organs (FitzGerald, 1988a; Ireland and Maisel, 1989; Merdes et al., 1991) and no observed hybridization of a labeled filensin probe to RNA from other organs (Masaki and Watanabe, 1992; Gounari et al., 1993), most reports have suggested that filensin is lens-specific. However, one study reported weak immunoreactivity of a monoclonal antibody to bovine lens 115 kDa protein with a 115 kDa component of urea-insoluble protein pellet fractions from bovine heart and red blood cells (FitzGerald and Casselman, 1990).

Anti-filensin antibodies labeled lens fiber cell membranes, but did not label lens epithelial cells (Fig. 7). The lens fiber cell specificity observed in this study was consistent with the reported localization of chicken CP95 (Ireland and Maisel, 1984, 1989), bovine 115 kDa (FitzGerald, 1988a) and porcine filensin (Merdes et al., 1991). The subcellular localization of filensin within the lens fiber cell has been reported to be membrane-associated and localized to the cytoplasmic lens beaded filaments. Immunofluorescence (this study; FitzGerald, 1988a; Ireland and Maisel, 1989; Merdes et al., 1991) and immunoelectron microscopy (Merdes et al., 1991) demonstrated a membrane association for filensin, while immunoelectron microscopy localized CP95 (Ireland and Maisel, 1989) and bovine 115 kDa (FitzGerald and Gottlieb, 1989) to the cytoplasmic lens beaded filament. The labeled beaded filaments may reside adjacent to the plasma membrane (Ireland and Maisel, 1989).

The deduced amino acid sequence of chicken filensin did not contain a predicted membrane spanning domain of hydrophobic amino acids, supporting cell fractionation studies that concluded that filensin was a peripheral membrane protein (Merdes et al., 1991; Brunkener and Georgatos, 1992). Chicken filensin also lacked amino-terminal and car-boxy-terminal consensus sequences for covalent lipid modifications, containing neither a penultimate amino-terminal glycine for myristylation nor a carboxy-terminal-CAAX for isoprenylation (Magee, 1990). The noted association of at least some filensin molecules with the plasma membrane could occur through the binding of filensin to an integral membrane protein (Brunkener and Georgatos, 1992).

Immunoblots of chicken lens proteins probed with antifilensin antibodies indicated that embryonic chicken filensin migrated in SDS-PAGE as a triplet of 102, 105 and 109 kDa, while adult rooster filensin migrated as a 105/109 kDa doublet and a broad band at 40 kDa. The isolation of different subsets of filensin protein bands from embryonic chicken lens and from adult rooster lens implied that the multiple protein forms were endogenous to the lenses and not due simply to protein degradation during preparation for gel electrophoresis. Post-translational modifications, including protein phosphorylation and/or protease cleavage, are possible explanations for the observation of multiple filensin protein bands from the lens.

CP95 of chicken has been reported to consist of a family of related protein bands with apparent molecular masses ranging from a 97/93 kDa doublet to 49 kDa (Ireland and Maisel, 1989). Incubation of chicken lenses in the presence of [³⁵S]methionine resulted in the initial synthesis of the 93 kDa form only and in the cortical lens fiber cells only (Ireland and Maisel, 1989). Incubation of chicken lenses in the presence of [³²P]orthophosphate yielded labeled 97 and 93 kDa forms in the cortical fiber cells, as well as labeled lower molecular mass forms in the older fiber cells of the lens nucleus (Ireland and Maisel, 1989). These data suggested that CP95 was post-translationally modified and degraded with time in maturing lens fiber cells.

Other investigators reported that porcine filensin migrated as a 110/100 kDa doublet and major breakdown products of 54/51 kDa (Merdes et al., 1991). Multiple breakdown products of bovine 115 kDa protein were suggested to occur in vivo and to reflect age-related changes (FitzGerald, 1988b). A 51 kDa filensin band, also observed in bovine lens membrane fractions and localized to the older fiber cells in the lens nucleus, was presumed to represent a naturally occurring breakdown product of filensin (Brunkener and Georgatos, 1992). The multiple chicken filensin forms observed in immunoblots in this study may represent endogenous protein processing.

The predicted amino acid sequence of chicken filensin contained a 280 amino acid putative rod domain (residues 39 to 318), dominated by a heptad repeat pattern in which the first and fourth amino acids of the repeat were hydrophobic. The filensin rod domain was punctuated by short interruptions in the heptad repeat pattern, and these interruptions aligned with the positions of linkers in the intermediate filament rod domain. Insertion of a four-heptad repeat gap in the chicken filensin sequence, corresponding to the amino-terminal half of intermediate filament rod coil 2, optimized the alignment of the shorter 280 amino acid rod domain of chicken filensin with the 310 amino acid rod domain of types I-IV intermediate filament proteins. A stutter in the heptad repeat phasing of chicken filensin occurred at amino acid 256, and exactly matched the stutter in the middle of intermediate filament rod coil 2 in phase shift and in position with relation to the end of the rod domain.

The amino acid sequence identity between the putative filensin rod and the rod domain of intermediate filament proteins was 20 to 25%, which is comparable to the sequence homology found among intermediate filament proteins of different types (Steinert and Roop, 1988). Filensin contained sequence attributes of intermediate filament proteins; however, the length of the putative filensin rod was shorter than that found in any intermediate filament protein sequenced to date.

Keratin intermediate filaments are obligate heteropolymers composed of equimolar amounts of acidic type I and neutral/basic type II monomers. Type I and type II keratin genes are expressed in defined pairs that are spatially (tissue specific) and temporally (developmentally) regulated. Recent reports have suggested that the epithelioid lens may also contain cytokeratins (Magin et al., 1990; Bader, B. L., Magin, T. M. and Franke, W. W. (1992) Abstract: Fifth International Congress on Cell Biology, p. 79).

FastDB comparison of the chicken filensin deduced amino acid sequence with all sequences in the data bases yielded higher initial alignment scores with type IIb cytokeratins than with other intermediate filament proteins. If filensin evolved from a keratin gene and functions as a lensspecific type II cytokeratin, then chicken lens CP49 may represent the filensin polymerization partner. Chicken CP49 is another lens fiber cell-specific, cytoskeletal protein which, along with filensin, has been immunolocalized to lens fiber cell beaded filaments (Ireland and Maisel, 1984). The corresponding 47 kDa mammalian lens protein bound to filensin in vitro (Merdes et al., 1991), and preliminary cDNA sequence analysis of bovine 47 kDa indicated sequence similarity to intermediate filament proteins, specifically to type I cytokeratins (A. Merdes and S. D. Georgatos, personal communication). The partial cDNA sequence of the mouse lens CP49 equivalent also showed sequence similarity to type I cytokeratins (Hess et el., 1993). A more detailed assessment of the relationship of filensin to the large intermediate filament family of proteins awaits the complete sequence analysis and protein biochemical characterization of CP49.

I thank R. G. Johnson for providing support for this research and an excellent working environment. I thank R. A. Meyer (University of Minnesota, St. Paul, MN, USA) for the immunofluorescence data and H. Maisel (Wayne State University School of Medicine, Detroit, MI, USA) for the gift of monoclonal antibody MA₂. I thank S. D. Georgatos, F. Gounari, A. Merdes and C. A. Ouzounis (European Molecular Biology Laboratory, Heidelberg, FRG), R. Quinlan (University of Dundee, Dundee, Scotland, UK), P. Fitzgerald and J. Hess (University of California Davis, Davis, CA, USA) for their suggestions during preparation of the manuscript and for sharing information prior to publication.

These data are based upon work supported under a National Science Foundation Graduate Fellowship. The work was supported in part by NIH grant GM 37230 awarded to R. G. Johnson.