Differential assembly of α- and γ-filagenins into thick filaments in : Caenorhabditis elegans

The assembly of myosin into thick filaments in skeletal muscles has provided a model system for investigating the assembly of biological structures (Huxley, 1963; Epstein and Fischman, 1991; Liu et al., 1997). Thick filaments are highly organized structures with lengths, diameters, and flexural rigidities characteristic of their muscle of origin (Craig, 1977; Kensler et al., 1985; Stewart and Kensler, 1986; Schmid and Epstein, 1998). Myosin molecules in different regions of thick filaments show distinct biochemical characteristics, including interactions with different myosin binding proteins. Thick filaments, therefore, are highly differentiated structures. To date, such differentiated assembly has not been achieved with purified myosins. How cells assemble such filaments, therefore, remains an important question (Epstein and Fischman, 1991; Barral and Epstein, 1999).

Accumulating evidence suggests that the assembly of myosin into thick filaments may require other proteins rather than being the result of simple self-assembly. The myosin binding protein C appears to be an integral part of thick filaments in vertebrate cardiac and skeletal muscles (Morimoto and Harrington, 1974), and its presence decreases the critical concentration for the assembly of purified myosin into synthetic filaments (Davis, 1988) and increases their rigidity (Schmid and Epstein, 1998). The giant protein titin may also contribute to the assembly of thick filaments in muscles (Labeit and Kolmerer, 1995). A major deletion within titin in the myofibroblast cell line BHK-21/C13 results in defective myofibril assembly and reduced thick filaments (van der Ven et al., 2000). In contrast, loss-of-function mutants in C. elegans twitchin and Drosophila D-titin, homologues of vertebrate titin, still assemble thick filaments (Waterston et al., 1980; Zhang et al., 2000). The existence of additional proteins in thick filaments is evident in invertebrates as paramyosin, a homologue of the myosin rod, is an integral part of thick filaments and forms their backbone (Cohen et al., 1971; Szent-Györgyi et al., 1971; Levine et al., 1983; Epstein et al., 1985). Other proteins that appear to be thick filament components include zeelins in Lethocerus muscles (Ferguson et al., 1994) and flightin in Drosophila flight muscle (Vigoreaux et al., 1998). In Drosophila, multiple myosin isoforms are expressed in structurally distinct thick filaments of different muscles. These characteristic thick filaments are retained in a genetically engineered Drosophila strain that expresses only a single myosin isoform, suggesting that factors other than myosin itself determine the structure in each muscle type (Wells et al., 1996).

C. elegans provides a useful model to study the complex assembly of thick filaments because of well-established biochemical, genetic and structural backgrounds. The three major components, myosin A, myosin B and paramyosin, of thick filaments from the body wall muscle of C. elegans are well-characterized biochemically (Waterston et al., 1974; Waterston et al., 1977; Harris and Epstein, 1977; Schachat et al., 1977). The genes encoding these three proteins have been sequenced (McLachlan and Karn, 1982; Karn et al., 1983; Kagawa et al., 1989). Myosins A and B are homodimers of distinct heavy chains, encoded by myo-3 and unc-54, respectively (Schachat et al., 1977; Schachat et al., 1978). Paramyosin is encoded by unc-15 (Waterston et al., 1977), and shares 38% identity in amino acid sequence with the rod domains of myosin heavy chains (Kagawa et al., 1989). Myosin A is localized in the center of the thick filament while myosin B is present only in the flanking regions (Miller et al., 1983). The paramyosin core of the thick filament still contains myosin A in its center and is proposed to serve as a template for myosin A and myosin B to assemble into their distinct regions to form the native thick filament (Epstein et al., 1985; Deitiker and Epstein, 1993). The function of myosin A appears to be required for either initiation or stabilization of thick filament assembly as myo-3 null mutants do not produce stable thick filaments at all, leading to embryonic lethality (Waterston, 1989). unc-15 null mutants assemble very thin, abnormal filaments, and the animals are paralyzed. Myosin B appears nonessential for thick filament assembly since myosin A assembles in place of myosin B in thick filaments of unc-54 mutants and myosin A overexpression suppresses myosin B defects (Epstein et al., 1986).

Synthetic filaments of purified C. elegans myosin and paramyosin do not show the characteristics of natural thick filaments isolated from muscles (Harris and Epstein, 1977). In vivo, other C. elegans gene products appear to affect thick filament assembly by interacting with myosin or paramyosin. For example, the unc-45 product has been shown to interact with myosin heavy chains encoded by myo-3 and unc-54 by genetic tests (Venolia and Waterston, 1990; Ao and Pilgrim, 2000). Both temperature-sensitive unc and lethal pat mutants of unc-45 affect the assembly of myosin into thick filaments (Epstein and Thomson, 1974; Barral et al., 1998). The unc-82 product interacts with paramyosin encoded by unc-15 as shown by genetic tests (Waterston et al., 1980), and unc-82 mutant structures are similar to the paracrystalline structures found in unc-15 missense mutants (Epstein et al., 1986).

Structural studies of thick filament cores dissociated from natural thick filaments have shown that they are composed of paramyosin sub-filaments that are crosslinked by additional proteins (Deitiker and Epstein, 1993; Epstein et al., 1995; Liu et al., 1998). Three proteins of 30, 28 and 20 kDa have been found to copurify with the cores (Deitiker and Epstein, 1993). One of them, the 28 kDa species named β-filagenin, coassembles with paramyosin or with myosin in the absence of paramyosin (Liu et al., 1998). Therefore, proper assembly of thick filaments in C. elegans may require interactions among multiple proteins. Here, we present evidence that the 30 kDa protein, now termed α-filagenin, and the 20 kDa protein, now termed γ-filagenin, like β-filagenin are components of the cores.

As an approach to understanding the function of the filagenins, multiple attempts have been made to detect deletion mutations in the genes encoding the filagenins by PCR-screening of mutagenized stocks in collaboration with the C. elegans Gene Knockout Project at the Oklahoma Medical Research Foundation. At least two sets of primers for each filagenin gene were used to screen a total of 7.25 million genomes in eleven distinct libraries. Mutagens used to develop those libraries include ethylmethanesulfonate, diepoxybutane, formaldehyde and ultraviolet/trimethylpsoralen. No stable deletions have been found in any of the genes encoding the filagenins. About one in seven genes behave in this manner (G. Moulder, personal communication). Surprisingly, multiple attempts to inhibit the expression of α-,β- and γ-filagenins with the RNA interference technique have failed to either create a phenotype or block the expression of the specific protein (A. Hutagalung, F. Liu, and H. F. Epstein, unpublished results). However, RNA interference may be rescued by significant expression of maternal messenger RNAs.

We have studied therefore the localization of these proteins at cellular, subcellular and ultrastructural levels as a function of developmental change, to provide possible insights about their function. Our results indicate that the filagenins are expressed sequentially, concomitant with the assembly of thick filaments of distinct lengths at specific developmental stages in C. elegans.

N2 (wild-type) C. elegans strain was grown at 20°C on a lawn of Escherichia coli strain NA22 seeded on the peptone-enriched agar plates (Brenner, 1974; Schachat et al., 1978). L4 larvae for thick filament purification were harvested and mixed with two volumes of OCT embedding medium and stored in liquid nitrogen (Epstein et al., 1988; Deitiker and Epstein, 1993). Gravid adults were used to isolate embryos for immunofluorescence microscopy.

The purification of thick filament proteins was performed as previously described (Epstein et al., 1988; Deitiker and Epstein, 1993; Liu et al., 1998). Thick filaments were first isolated and purified to the 15,000 g pellet stage. Thick filaments in the 15,000 g pellet were further purified by gradient centrifugation using 19%-38% sucrose. Highly purified but diluted thick filaments in the fractions of the gradient were concentrated by ethanol precipitation (Harris and Epstein, 1977). Thick filament proteins were then separated by 11% SDS-PAGE.

To obtain the amino acid sequence of α-filagenin, thick filament proteins separated on 11% SDS-PAGE were transferred to a polyvinylidene difluoride (PVDF) membrane and stained with Coomassie Blue. The membrane region containing α-filagenin was excised, and the protein was eluted from the membrane and digested with endoprotease Lys-C (Promega, Madison, WI, USA). For γ-filagenin, thick filament proteins were also separated by 11% SDS-PAGE, but not transferred to PVDF membrane. Instead, the γ-filagenin-containing band was directly excised from the gel, and the protein was eluted and digested with trypsin (Promega, Madison, WI, USA). Peptide fragments of α- and γ-filagenins were separated by high-performance liquid chromatography and sequenced on the model 477A protein sequencer (Applied Biosystems, Foster City, CA, USA).

The cDNA of α-filagenin was generated by RT-PCR using total RNAs isolated from nematodes of mixed stages. The PCR primers used were 5′ CCC CGG GTC GAC CAT GTC AAC CAT TGC GGC TGG 3′ and 5′ CCC CGG GTC GAC TTA GAA GAA GAA ACG GCC GGT ACG 3′. A single band of 900 bp was generated, subcloned and sequenced. For γ-filagenin, two PCR primers, 5′ GAG GGA TCC AGA TAC TAC CAT CTC GC 3′ and 5′ CGG GGT ACC GGT TGG TTG GAA TAA GTG GGC GGA G 3′, based on partial matches of the genomic sequences and the available peptide sequences of γ-filagenin from proteolytic fragments of the protein, were used to amplify a C. elegans cDNA library. A 400 bp PCR fragment was generated and used as probe to screen the same cDNA library. Multiple cDNA clones were obtained and sequenced.

C. elegans sequences were obtained from the C. elegans Genome Sequencing Project (The Sanger Centre, Hinxton Hall, Cambridge, UK and the Washington University School of Medicine, St Louis, MO, USA), and from the C. elegans EST database (http://www.ddbj.nig.ac.jp/). Sequence similarity searches were performed using the NCBI Blast search engine (http://www.ncbi.nlm.nih.gov/BLAST/). Multiple sequence alignments were done using the Baylor College of Medicine Search Launcher (Smith et al., 1996). Secondary structure predictions used the Garnier-Osguthorpe-Robson method (Garnier et al., 1978).

For α- and γ-filagenins, antibody production followed the procedures described previously (Liu et al., 1998). Peptides DVDQDTKWWKEYPS from the partial amino acid sequence of α-filagenin and FLLRPLIPTNLVR from the partial amino acid sequence of γ-filagenin were separately synthesized onto a multiple antigenic peptide (Tam, 1988). The peptides were chosen based on their predicted high accessibility and antigenicity using the Surface Probability and Antigenicity programs in the GCG Package, Wisconsin Package Version 9.0, GCG, Madison, Wisconsin, USA (http://mbcr.bcm.tmc.edu/Guide/GCG.html). These peptides were used to immunize 5-6 month-old female white New Zealand rabbits using complete Freund’s Adjuvant for initial inoculations, followed with incomplete Freund’s Adjuvant in the following boosts (Cocalico Biologicals, Inc., Reamstown, PA, USA). The same peptides were conjugated to a gel matrix (Affi-gel 10, Bio-Rad Laboratories, Hercules, CA, USA) and used to affinity-purify the antiserums. Affinity-purified antibodies were characterized for their specificity by western blot using nematode total protein.

For whole mounts of adult nematodes, the procedure by G. Ruvkun (Miller and Shakes, 1995) was followed. Nematodes were cracked open by three rounds of freezing and thawing in 80 mM KCl, 20 mM NaCl, 10 mM Na₂EGTA, 5 mM spermidine-HCl, 15 mM sodium 1, 4-piperazinediethanesulfonic acid (Pipes), pH 7.4, and 25% methanol, followed by 30 minutes of fixation in 3% freshly made paraformaldehyde. After processing as described in the protocol, nematodes were incubated overnight at room temperature with a mixture of a rabbit polyclonal antibody against either α- or γ-filagenin, each at 5 μg/ml, and a mouse monoclonal antibody against either myosin heavy chain A or B, each at 1 μg/ml. After washing for 4-8 hours, nematodes were reacted with a mixture of goat anti-rabbit IgG conjugated with the cyanine dye Cy3 and goat anti-mouse IgG conjugated with the cyanine dye Cy2 (Jackson ImmunoResearch Laboratories, West Grove, PA, USA), at dilutions of 1:2,000-3,000, overnight at room temperature. After 4-8 hours of washing, nematodes were mounted on glass slides with 5% n-propyl gallate in glycerol. Specimens were viewed using the Olympus BX60 model microscope (Olympus Optical Co., Japan). Photomicrographs were taken using Provia Fujichrome 1600 Professional film (Fuji Photo Co., Japan).

Gravid adults were used for the isolation of embryos. After dissolving nematodes with 1.25% sodium hypochlorite, 0.5 M NaOH, embryos were rinsed with M9 buffer three times. Embryos were then fixed for 20 minutes at room temperature with freshly made 3% paraformaldehyde in phosphate-buffered saline (PBS, 137 mM NaCl, 3 mM KCl, 10 mM Na₂HPO₄, 2 mM KH₂PO₄, pH 7.2). Embryos were further rinsed with PBS three times, 10 minutes each, and then transferred to −20°C methanol. Embryos were generally stored in methanol at −80°C before use. Prior to antibody staining, embryos were first rehydrated with descending concentrations of methanol at 75%, 50%, 25% and PBS twice, 10 minutes each, and then freeze-fractured as described previously (Epstein et al., 1993). Embryos were incubated with 100% goat serum for 1 hour at 37°C to block nonspecific binding before reacting with antibodies. All antibodies were diluted into 50% goat serum. Rabbit polyclonal antibodies against α-filagenin or γ-filagenin (5 μg/ml each) were respectively mixed together with mouse mAb 5-6 against myosin heavy chain A (1 μg/ml), and reacted with embryos for 1 hour at 37°C. After three washes with PBS, embryos were reacted with affinity-purified goat anti-rabbit IgG conjugated with Cy3 and affinity-purified goat anti-mouse IgG conjugated with Cy2 (1:2,000-3,000 dilution). Embryos were mounted after three washes with PBS and photographed as described for adult nematodes in the previous section.

Thick filaments were purified to the 15,000 g step to reduce the amount of actin filaments (Deitiker and Epstein, 1993). The 15,000 g pellets were resuspended in filament isolation buffer and dissociated on ice for 30 minutes after adding NaCl to 400 mM. Solubilized proteins were separated from the remaining structures (cores) by centrifugation at 100,000 g for 40 minutes with the TL-100 ultracentrifuge using the microcentrifuge PA rotor (Beckman Instruments, Inc., Palo Alto, CA, USA). The cores were brought to equal volume with the supernatant. Equal volumes of thick filaments (resuspended 15,000 g pellet), solubilized proteins and cores (resuspended pellet after dissociation) were subjected to 11% SDS-PAGE, and separated proteins were western blotted with antibodies against α- and γ-filagenins.

Procedures for electron microscopy were as described (Miller et al., 1983; Epstein et al., 1985; Liu et al., 1998). Antibodies against paramyosin, α-filagenin and γ-filagenin were all used at 50 μg/ml. Affinity-purified goat antibodies against mouse or rabbit IgG were used at 20 μg/ml as secondary antibodies to enhance detection. Measurements of periodicities of antibody reactions were done using the EPOI LP-6 Profile Projector (Ehrenreich Photo-Optical Industries, Inc., Garden City, NY, USA) fitted with micrometers (The L. S. Starrett Co., Athol, MA, USA). To avoid potential systematic errors that may be caused by the electron microscope during experiments, micrographs of α- and γ-filagenin reactions with core structures were taken in the same session as calibration standards of bovine catalase crystals and tobacco mosaic virus. Grid preparation for both samples took place at the same time. In addition, data for each reaction were obtained from multiple grids of the same sample in each session. Lens currents were not normalized for these sessions; however, multiple grids of the different samples were examined alternately to minimize any artifactual differences. Differences between antibody labeling patterns of α- and γ-filagenins were analyzed by the t-test for populations with homogeneous variances.

Previously, three proteins with apparent molecular masses of 30, 28 and 20 kDa were found to cosediment with the tubular cores of C. elegans body wall muscle thick filaments (Deitiker and Epstein, 1993). Characterization of the 28 kDa protein has been reported (Liu et al., 1998). This protein coassembles with either paramyosin of the tubular wild-type filament cores or with myosin B of CB1214 paramyosin-deficient mutant core-like filaments. It was named β-filagenin because neither purified myosin nor paramyosin assemble into tubular structures (Harris and Epstein, 1977). The 30 and 20 kDa proteins have been termed α-filagenin and γ-filagenin because they also copurify with thick filament cores, like β-filagenin, and share multiple properties with β-filagenin with respect to thick filament structure and assembly.

Partial amino acid sequences of α- and γ-filagenins were determined by gas-phase Edman degradation of proteolytic fragments of the two proteins from purified thick filaments. Two sequences, SYSDARWRDTHREINK and RIFFDEK, were obtained from α-filagenin and used to search the C. elegans genome database (http://genome.wustl.edu/gsc/ index.shtml; http://www.sanger.ac.uk/Projects/C_elegans/). These two sequences matched perfectly with two sequences from a predicted protein of molecular mass of 30,043 Da (250 amino acids) that was assembled from open reading frames within a single cosmid (C46G7.2. GenBank accession number U97593) containing C. elegans genomic DNA located in the central region of chromosome IV (Fig. 1A). The open reading frame within this gene was verified by partial cDNA sequences from C. elegans EST database (http://www.ddbj.nig.ac.jp/) and by sequencing a 900 bp cDNA product amplified by RT-PCR from C. elegans total RNA. Similar to β-filagenin, α-filagenin is predicted to be basic with an isoelectric point of 10.52. In contrast to β-filagenin, which is predicted to have β-strands only, α-filagenin is predicted to contain both α-helices and β-strands.

Two peptide sequences, YYHLA and FLLRPLIPTNLVR, were obtained from proteolytic peptide fragments of γ-filagenin. When these were used as queries to search the C. elegans genome database, short stretches of matches were found in a single cosmid but in different open reading frames. Two PCR primers, 5′ GAG GGA TCC AGA TAC TAC CAT CTC GC 3′ and 5′ CGG GGT ACC GGT TGG TTG GAA TAA GTG GGC GGA G 3′, based on partial matches of the genomic sequences and the available peptide sequences of γ-filagenin from proteolytic fragments of the protein, were used to amplify a C. elegans cDNA library. A 400 bp PCR fragment was generated and used as probe to screen the same cDNA library. Multiple cDNA clones were obtained and a cDNA encoding the full length of γ-filagenin was sequenced. The gene encoding γ-filagenin was located within the cosmid F42G4 (GenBank accession number Z81082), which maps to the right arm of chromosome II. The predicted γ-filagenin contains 166 amino acids and has a calculated molecular mass of 19,601 Da (Fig. 1B). γ-filagenin is predicted to be very basic with an isoelectric point of 11.49 and, like β-filagenin, to contain only β-strands as regular secondary structures.

Multiple homologues of α- and γ-filagenins are found with high conservation at the amino acid level in various nematodes (Fig. 2). For example, the respective identities of the homologues with C. elegans α-filagenin (measured within respective length of each protein) are: Anisakis simplex (GenBank accession number, AJ250666), 47.8%; Brugia malayi (AA406674), 51.3%; Onchocerca volvulus (AI205475), 54.7%; Zeldia punctata (AW773354), 50.8%. The identities with the C. elegans γ-filagenin are Acanthocheilonema viteae (U07024), 55.3%; Ascaris suum (AW165735), 47.6%; Brugia malayi (AW225343), 56.8%; Onchocerca volvulus (U14530), 60.2%. Ascaris suum and Anisakis simplex belong to the order Ascaridida while Acanthocheilonema viteae, Brugia malayi and Onchocerca volvulus belong to the order Spirurida. Ascaridida and Spirurida diverged from Rhabditida (the order including C. elegans and Zeldia punctata) about 550 million years ago (Poinar, 1983). We have not found homologues of significant identity or similarity in amino acid sequence to either α- and γ-filagenins in Drosophila or vertebrate databases. However, proteins such as vertebrate myosin-binding C protein family members may share similar functions with the filagenins although they show no primary sequence homology.

In order to verify that α- and γ-filagenins are core components of C. elegans body wall thick filaments, specific antibodies against selected peptides (underlined sequences in Fig. 1) from the two proteins were produced. These sequences were chosen because they were predicted to be highly accessible and antigenic using the Peptidestructure programs from GCG. For all experiments, antibodies against the filagenins were affinity-purified using the peptide antigens as ligand and tested for specificity. Both affinity-purified antibodies showed reactions with a single band with nematode homogenates on western blots. Fig. 3 shows verification that α- and γ-filagenins are associated with cores. In these experiments, samples highly enriched in thick filaments (15,000 g pellet) were first dissociated with 400 mM NaCl to generate cores, and the solubilized proteins and cores were then separated by centrifugation. The cores (pellet) were brought to equal volume with the supernatant. Proteins from equal volumes of thick filaments (15,000 g pellet), solubilized proteins from the dissociation of 15,000 g pellet, and cores (pellet after dissociation) were then separated by 11% SDS-PAGE. Coomassie Blue staining of the SDS-PAGE showed that thick filaments contained mainly myosin and paramyosin, although there was some actin contamination (Fig. 3, panel CB, lane T). Most myosin and paramyosin were dissociated from the thick filaments (Fig. 3, panel CB, lane S), but a certain amount of paramyosin remained with the cores (Fig. 3, panel CB, lane P). Simultaneous western blotting of thick filaments, solubilized proteins and the cores showed that α- and γ-filagenins were exclusively associated with thick filaments and cores, but not with the solubilized proteins (Fig. 3, panels α and γ).

α- and γ-filagenins were further verified as core components by immunoelectron microscopy of isolated cores using specific antibodies. Isolated, intact thick filaments do not react with these antibodies; only the dissociated cores do. Fig. 4 shows cores labeled with antibodies against α-(Fig. 4A,B) and γ-filagenins (Fig. 4C,D), respectively, followed by reacting with a goat anti-rabbit IgG secondary antibody. The reactions (arrowheads) revealed apparent repeating patterns for both antibodies. The large complexes of the antibody reactions were consistent with those obtained with paramyosin (Deitiker and Epstein, 1993) and β-filagenin (Liu et al., 1998), when the antigenic sites are well dispersed as here (>70 nm). This is in contrast to myosin, where the sites are only 14.5 nm apart and there is continuous labeling (Miller et al., 1983; Epstein et al., 1986). With the more disperse antigens, a larger complex of primary and secondary antibodies develops above each site, in contrast to the more extensive crosslinking of antigens that can be bridged by immunoglobulin molecules.

Measurements of the labeling of cores with antibodies against α- and γ-filagenins showed that the repeats of the specific antibody reactions with the two filagenins were distinct. The mean periodicities of the anti-α-filagenin and anti-γ-filagenin localizations were measured to be 74.8±3.2 nm (n=48) and 71.9±3.8 nm (n=286), respectively. The difference between the means is significant (P=0.0018). Anti-β-filagenin localizations yield a periodicity of 74.7±2.1 nm (n=29) (Liu et al., 1998; F. Liu, I. Ortiz, and H. F. Epstein, unpublished results).

Careful examination of the core regions reacting with respective anti-filagenin antibodies also revealed differences in their diameters. The α-filagenin-reacting core regions appeared larger in diameter than the γ-filagenin-reacting regions (Fig. 4). Since the medial region of the core is larger in diameter than the flanking regions (Epstein et al., 1985), the two filagenins may be differentially localized, with α-filagenin in the center and γ-filagenin in the flanking regions.

The thick filaments of C. elegans body wall muscle contain two myosin isoforms, each with distinct localization. Myosin A is restricted to the center of the thick filament while myosin B is localized in the flanking regions (Miller et al., 1983). By immunofluorescence, myosin A labeling is observed as narrow lines in the middle of the thick filament-containing A-bands (arrowheads in Fig. 5B,F), and myosin B is labeled throughout each A-band except for the medial gap (arrows in Fig. 5D,H). To determine whether the filagenins are differentially localized, whole mounts of adult C. elegans were examined by immunofluorescence microscopy with double labeling reactions using affinity-purified rabbit polyclonal antibodies against each filagenin and mouse monoclonal antibodies against myosin isoforms A or B.

Fig. 5A-D shows the double labeling of C. elegans body wall muscle cells with affinity-purified rabbit polyclonal antibody against α-filagenin and mouse mAbs against myosin A or B. Myosin A was labeled with its typically restricted localization in the center of each A-band (arrowhead in Fig. 5B). α-filagenin clearly overlapped with myosin A, with an apparently slightly wider distribution (arrowhead in Fig. 5A). Comparison of α-filagenin (arrowhead in Fig. 5C) with myosin B in the flanking regions (arrows in Fig. 5D) verified that α-filagenin is localized in the center of the A-band.

Fig. 5E-H shows the double labeling of γ-filagenin and myosin isoforms of C. elegans body wall muscle cells. In contrast to α-filagenin, γ-filagenin (arrows in Fig. 5G) colocalized with the flanking myosin B (arrows in Fig. 5H), but not with the centrally localized myosin A (compare arrowheads in Fig. 5E,F). These results obtained by immunofluorescence microscopy confirmed the suggestion from electron microscopy (Fig. 4) that α- and γ-filagenins may be localized to distinct regions along the length of the thick filament. α- and γ-filagenins were also localized in anal and vulval muscles, but not detected in pharyngeal muscle (data not shown).

In vivo thick filament assembly in C. elegans appears to be a regulated developmental process that involves initiation, growth and determination of the lengths of new structures. Thick filaments have distinct lengths in embryos (Gossett et al, 1982), larvae (Waterston and Francis, 1985) and adults (Mackenzie and Epstein, 1980). To investigate what role filagenins may play in these processes, we compared the expression of α- and γ-filagenins during development.

Fig. 6A shows a 1.5-fold embryo labeled with the anti-myosin A antibody. A-bands, which contain thick filaments, start to form after this stage as the linear structures align along the body axis of the embryo (arrowheads; Epstein et al., 1993). The expression of α-filagenin was prominent and clearly colocalized with myosin A in the thick filament-containing A-bands in the same embryo (arrowheads in Fig. 6A,B). Embryos of similar stage expressing assembled myosin A (arrowheads in Fig. 6C) did not express detectable levels of γ-filagenin (Fig. 6D).

Fig. 7 shows the expression of α- and γ-filagenins in embryos at stages when myofibrils are formed and enlarge. Fig. 7A shows a threefold embryo labeled with the mAb against myosin A. At this stage myosin A was well assembled into A-bands (arrowheads). Fig. 7B shows the expression of α-filagenin in the same embryo. Overlapping of myosin A and α-filagenin localizations in A-bands was clear (compare arrowheads in Fig. 7A,B). However, γ-filagenin was not detected, as shown in an embryo of similar stage (Fig. 7C,D). Fig. 6E-H shows two vermiform stage embryos just prior to hatching. In both embryos, A-bands were well-organized, as shown by antibody labeling of myosin A (arrowheads in Fig. 7E,G). At this stage γ-filagenin (Fig. 7H) as well as α-filagenin (Fig. 7F) was expressed and assembled into A-bands. Myosin A clearly overlapped with α- and γ-filagenins (compare arrowheads in Fig. 7E-H). β-filagenin was not detected during embryogenesis.

The expression and assembly of α-, γ- and β-filagenins appear to correlate with the formation of A-bands of increasingly larger sizes and, therefore, thick filaments of increasing lengths. By electron and polarized light microscopy, it was shown that thick filaments in A-bands in adult body wall muscle do not stagger (Mackenzie et al., 1978). The angle between thick filaments and the oblique striations of the sarcomere was determined to be 5.9°. Therefore, the lengths of thick filaments in A-bands can be calculated by using the sine law if the widths of A-bands are known (Mackenzie and Epstein, 1980). Since the width of the A-band in the body wall muscle of adult C. elegans was measured by polarized light microscopy to be 1.0 μm, the length of the thick filament in adults was calculated to be 9.7 μm accordingly. Based on the same assumptions, the length of the thick filaments in L1 larvae was estimated to be about 5.0 μm since the width of their A-bands was about half that of adults (Waterston and Francis, 1985). We assume that thick filaments in A-bands in early L1 larvae and embryos have the same oblique angles and do not stagger. By obtaining the widths of their A-bands, we should be able to calculate the respective lengths of their thick filaments. However, it is difficult to measure the widths of A-bands in early L1 larvae and embryos using polarized light microscopy because of the very small amounts of birefringence present. Therefore, we used immunofluorescence as an alternative method.

Fig. 8 shows the labeling of A-bands in muscles of a late embryo (Fig. 8A), an L1 larva (Fig. 8B) and an adult (Fig. 8C) with a mouse mAb against myosin B followed by a fluorescent secondary goat antibody against mouse IgG. We know that each of the myosin B-labeled structures indicated in Fig. 8A corresponds to a myosin A-containing A-band in Fig. 7E,G. The paired structures of the larval muscle cells indicated in Fig. 8B have been shown previously by electron microscopy to be proper A-bands (Mackenzie et al., 1978). Because of the divergence of fluorescence after its emission from its original source, the measured widths of the A-bands were larger than their real sizes. For example, the A-bands of adults were measured by immunofluorescence to be 1.7 μm whereas they are known to be 1.0 μm by polarized light and electron microscopy (Mackenzie and Epstein, 1980). The calculated ratio of the widths from the A-bands of late embryos:L1 larvae:adults as measured (n=5) for each stage by immunofluorescence was 0.2:0.48:1.00. After normalization against the known width of 1.0 μm of adult A-bands, the respective widths of the A-bands for embryos and larvae were calculated to be 0.2 μm and 0.48 μm, proportionally. According to the model of Mackenzie and Epstein (1980), the length of L1 larval thick filaments was calculated to be 4.7 μm, in agreement with Waterston and Francis (1985). Similarly, the length of thick filaments in the embryos was calculated to be 2.0 μm. Therefore, detection of the filagenins in A-bands appeared to be concomitant with the assembly of thick filaments of distinct lengths.

α- and γ-filagenins are novel proteins in the thick filament cores of C. elegans body wall muscle. Based on partial peptide sequences derived from purified proteins, we have cloned the cDNAs encoding the two proteins. Their amino acid sequences indicate that both α- and γ-filagenins, like the previously described β-filagenin, are highly basic proteins. Proteins with 47-60% identities with α- and γ-filagenins at the amino acid level are found in disease-causing nematodes including Brugia malayi and Onchocerca volvulus (humans), Ascaris suum (pigs), Acanthocheilonema viteae (rodents) and Anisakis simplex (fish), as well as in free-living Zeldia punctata. Those parasitic species belong to the orders Ascaridida and Spirurida that diverged from Rhabditida (which includes Zeldia punctata and C. elegans) about 550 million years ago (Poinar, 1983). Such high conservation in a span of a half billion years suggests important functions for the filagenins. We found that α- and γ-filagenins are localized to separate regions of thick filaments. By antibody labeling of isolated cores, we discovered that both filagenins show repeating patterns, but the periodicities are different. α-filagenin appears early in myogenesis whereas γ-filagenin appears just before hatching, suggesting that their expression is developmentally regulated. The apparent affinities of the antibodies against α- and γ-filagenins appeared similar (data not shown); therefore, they should detect comparable levels of expression of α- and γ-filagenins.

α-filagenin is localized in the medial region of cores, whereas γ-filagenin is localized in the flanking regions (Figs 4, 5). This result is similar, although not identical, to the respective localization of myosins A and B in thick filaments (Miller et al., 1983). An analogous situation may be found in vertebrate cardiac and skeletal muscle thick filaments. Members of the myosin binding protein-C family appear to be intrinsic proteins restricted to specific regions of cardiac and skeletal muscle thick filaments (Morimoto and Harrington, 1974; Moos et al., 1975; Einheber and Fischman, 1990; Okagaki et al., 1993). Copolymerization of myosin with C-protein consistently favors the formation of longer synthetic filaments with native diameters (Davis, 1988). Such synthetic thick filaments also show greater rigidity than myosin-only synthetic filaments (Schmid and Epstein, 1998). Coexpression of C-protein with myosin heavy chain in COS cells appears to promote the formation of myosin cables (Seiler et al., 1996). Other vertebrate proteins such as M-creatine kinase (Turner et al., 1973), M-band protein (Masaki and Takaiti, 1974) and myomesin (Strehler et al., 1980; Grove et al., 1985) are localized to the central regions of thick filaments.

In invertebrates including C. elegans, paramyosin is a major component of thick filament cores (Szent-Györgyi et al., 1971; Epstein et al., 1985). In the case of C. elegans, paramyosin forms subfilaments that are coupled by additional proteins into tubular cores (Deitiker and Epstein, 1993; Epstein et al., 1995; Liu et al., 1998). Filagenins have been proposed to be these coupling proteins because of their coassembly with paramyosin in the cores. The cores have been proposed to act as structural templates for myosin assembly (Epstein et al., 1985). Therefore, the distinct localizations of α- and γ-filagenins in cores may be related to the assembly and stabilization of myosins A and B in their respective regions in thick filaments. Furthermore, thick filaments show significant differences along their lengths. Their diameters vary from 33.4 nm in the center to 14.0 nm near the ends (Epstein et al., 1985). The mass per length of thick filaments and cores also decreases from the center to the flanking regions (Müller, S., Häner, M., Ortiz, I., Aebi, U. and Epstein, H. F., unpublished). The differential localizations of the filagenins and their different periodicities may contribute to such features of thick filaments, which may be critical in determining the precise length and dimension of native thick filaments.

Myosin assembly in C. elegans body wall muscle is a developmentally regulated process. As A-bands grow, the maximal length of thick filaments continues to increase from embryos to adults as well. Based on the model of Mackenzie and Epstein (1980) for measuring the length of adult thick filaments, the width of A-bands measured with immunofluorescence microscopy and their calibration to known lengths of adult filaments, the length of thick filaments in embryos was calculated to be about 2.0 μm (Fig. 8A; Gossett et al., 1982; Epstein et al., 1993). Similarly, the lengths of the body wall muscle thick filaments in L1 larvae were calculated to be 4.7 μm (Fig. 8B), consistent with the previously estimated value of 5.0 μm (Waterston and Francis, 1985). Therefore, these early thick filaments are significantly different from each other in size and are shorter than the 9.7 μm long thick filaments in adult body wall muscle (Fig. 8C; Mackenzie and Epstein, 1980).

We have found that the expression and assembly of specific filagenins correspond to distinct stages of muscle development in C. elegans. α-filagenin becomes closely overlapping with myosin A when myofibrils are formed in twofold embryos. In contrast, γ-filagenin is not detected until the late vermiform stage when myofibrils are well formed (Fig. 7). β-filagenin is detected only in larvae and adults (Liu, F. and Epstein, H. F., unpublished results).

Our results are consistent with a model in which individual filagenins may play sequential roles in the assembly of thick filaments with different lengths at specific developmental stages in C. elegans myogenesis (Fig. 9). In this model, α-filagenin would be involved in the assembly of short thick filaments of early embryonic myofibrils. α- and γ-filagenins both would act in the assembly of intermediate length thick filaments in more advanced, vermiform embryos and early larvae. α-, γ- and β-filagenins would all contribute to the assembly of long thick filaments during later larval and adult development.

We thank Dr M. W. Epstein for help with statistical analysis and Drs J. M. Barral and M. G. Price for discussions. Supported by grants from the Muscular Dystrophy Association, National Science Foundation, National Institutes of Health and National Space Biomedical Research Institute (NASA) to H.F.E., the Paul Cohen Neuromuscular Disease Research Fellowship from the MDA to F.L. and the NIH Research Service Award Postdoctoral Fellowship to C.C.B.