FER-1 regulates Ca²⁺-mediated membrane fusion during C. elegans spermatogenesis

Regulated exocytosis, whereby specialized vesicles are signaled to fuse with the plasma membrane (PM), is a necessary step for many cellular processes. These include the relay of neuronal signals, the acrosome reaction during sperm maturation, and the repair of torn membranes in a variety of cell types (reviewed by Gerasimenko et al., 2001; McNeil and Steinhardt, 2003). Although different from the acrosome reaction, the sessile spermatids of C. elegans must fuse multiple vesicles, called membranous organelles (MOs), to their PMs in order to mature into crawling spermatozoa capable of fertilization. This fusion event contributes new membrane and proteins to the PM (Chatterjee et al., 2005; Roberts et al., 1986; Ward et al., 1981; Xu and Sternberg, 2003). MOs are specialized, ER-derived vesicles that have a bi-lobed structure composed of a smaller head that is separated by an electron dense collar from a larger body. In fer-1 mutants, MOs do not fuse during spermatozoon maturation, even though MO heads abut the PM as if they are docked in preparation for fusion. A short pseudopod forms on fer-1 mutant spermatozoa, which moves ineffectively resulting in non-motile spermatozoa. Consequently, fer-1 spermatozoa cannot adhere to the uterine walls and are swept out of the spermatheca in hermaphrodites by the passage of oocytes (Ward et al., 1981; Ward and Miwa, 1978; Ward et al., 1982).

When fer-1 was identified and sequenced, no other proteins had strong resemblance (Achanzar and Ward, 1997). Subsequently, homologs were found by sequence similarity in humans and mice, forming a family of similar proteins now called `ferlins' (Bashir et al., 1998). Two of the human homologs, otoferlin and dysferlin, have mutations that cause human diseases. Mutations in otoferlin lead to nonsyndromic deafness DFNB9 (Yasunaga et al., 2000; Yasunaga et al., 1999) and mutations in dysferlin result in the autosomal recessive muscle diseases limb girdle muscular dystrophy 2B (LGMD2B) and Miyoshi myopathy (Bashir et al., 1998; Liu et al., 1998). Similar to the fer-1 mutant phenotype, LGMD2B/MM patients display defects in vesicle fusion in their muscle tissue. This results in an accumulation of vesicles below the sarcolemma during membrane repair after mechanical stress (Cenacchi et al., 2005; Selcen et al., 2001). Immunolocalization studies place normal dysferlin in the sarcolemma and in cytoplasmic vesicles. The PM localization is eliminated in LGMD2B/MM patients and cytoplasmic vesicle localization is only occasionally observed (Anderson et al., 1999; Bansal et al., 2003; Piccolo et al., 2000).

The ferlin family proteins are characterized by multiple C2 domains, a C-terminal transmembrane domain, and most have nested repeat sequences, termed Dysf-N and Dysf-C, of unknown function. C2 domains characterized in other proteins such as synaptotagmin, protein kinase C, and phospholipases, have been found to bind Ca²⁺, phospholipids and phosphotyrosine (for reviews, see Bai and Chapman, 2004; Nalefski and Falke, 1996; Rizo and Sudhof, 1998; Sondermann and Kuriyan, 2005). Additionally, C2 domains can mediate protein-protein interactions. Some of these interactions depend on Ca²⁺ whereas others do not require Ca²⁺ (Chapman et al., 1996; Chapman et al., 1995; Davis et al., 1999; Fukuda and Mikoshiba, 2000; Rickman and Davletov, 2003; Sutton et al., 1999).

The two C2 domains of synaptotagmin have been the most thoroughly studied and are thought to be the Ca²⁺ sensors for many types of regulated exocytosis. These C2 domains can interact with phospholipids in a Ca²⁺-dependent manner, and can also interact directly with other proteins of the membrane fusion machinery such as synaptobrevin and syntaxin (Chapman et al., 1995; Gerona et al., 2000; Rickman and Davletov, 2003). These interactions promote vesicle exocytosis in neurons, and lysosomal exocytosis during PM repair of fibroblasts (Brose et al., 1992; Rao et al., 2004; Reddy et al., 2001). Vesicle fusion during membrane repair in muscle tissue requires dysferlin and is triggered by a local influx of extracellular Ca²⁺ near sites of membrane disruption (Bansal et al., 2003). This suggests a model in which dysferlin might facilitate membrane repair by responding directly to a Ca²⁺ signal to trigger vesicle fusion (Bansal and Campbell, 2004). The protein sequence similarities and the similarity of cellular mutant phenotypes between dysferlin and FER-1 suggests a common role for these proteins during cell-type-specific membrane fusion events.

C. elegans sperm provide an excellent model to study the mechanisms involved in ferlin-regulated membrane fusion. C. elegans is easily manipulated genetically, and its spermatids can be readily isolated for in vitro examination and biochemical analysis. The MOs of spermatids accumulate below the PM and only fuse during spermiogenesis when the spermatids are activated to form spermatozoa in response to an in vivo or in vitro chemical signal. Thus, MO fusion is readily manipulated, allowing analysis of fusion in real time. Also, since spermatozoa activated in vitro retain fertility, as demonstrated by artificial insemination, the in vitro process mimics normal spermiogenesis (LaMunyon and Ward, 1994).

We have investigated the role of FER-1 in MO fusion using three approaches. First, we identified and characterized the phenotypes of ten fer-1 mutations. Second, we found that MO fusion is sensitive to intracellular calcium, particularly in fer-1 C2 domain mutants. Third, we showed that there are three isoforms of FER-1 in sperm and these are localized to the MO membranes.

Five of the mutations in fer-1 were identified in the initial cloning of the gene before its human homologs and functional domains were recognized (Achanzar and Ward, 1997). We have sequenced the remaining five fer-1 mutant alleles, for a total of six temperature-sensitive and four non-conditional alleles, each having a single base-pair mutation in fer-1 that alters the encoded protein. Table 1 lists, and Fig. 1 shows, the location of the predicted protein sequence changes caused by each mutation in relation to the conserved domains predicted for several ferlin family members. Two mutants introduce early stop codons. Seven of the eight missense mutations fall within predicted C2 or DYSF domains. The other missense mutation, hc80, is located between the last two C2 domains in a region of eight consecutive amino acid identities amongst the different homologs, which suggests another important functional region. It is striking that seven of the eight missense mutations fall into predicted domains, suggesting that these regions are more sensitive to amino acid substitutions than the rest of FER-1 and that each domain has a distinct and essential function.

That each of the C2 domains of FER-1 has distinct functions is supported by sequence analysis. A BLAST search (Altschul et al., 1990) with FER-1 identified one or more homologs in every animal genome that has been sequenced, but a striking absence in fungi and plants (Fig. 1). All ferlin proteins are characterized by the presence of multiple C2 domains, followed by a C-terminal transmembrane domain. To better understand the relationship between the C2 domains found in ferlins and in other well-studied proteins, we combined the individual ferlin C2 domain sequences with the 64 non-ferlin C2 domains aligned by Nalfeski and Falke (Nalfeski and Falke, 1996). From this new alignment we constructed a phylogenetic tree of these C2 domains. Fig. 2 illustrates a summary tree with the ferlin C2 domain branches color coded (as in Fig. 1) to reflect their position in the ferlin proteins (see supplementary material Fig. S2 for the full tree). Each colored branch in the tree represents individual ferlin C2 domains, which fall into distinct clades based on their color. This shows for all ferlins that a given C2 domain is more similar to others at a similar position in the protein than it is to the other C2 domains within the same protein. This strengthens the argument that each C2 domain has a distinct function, and suggests that this function has been preserved through ferlin protein evolution.

All fer-1 mutants have a sterile phenotype, but only the hc1ts and hc24ts mutants had been previously examined by electron microscopy and found to have temperature-sensitive defects in MO fusion (Ward et al., 1981). To see if fer-1 mutations in different domains caused distinguishable phenotypes, we developed a fast quantitative assay to examine MO fusion in live cells. When applied to spermatids, the lipophilic dye FM 1-43 partitions into the outer membrane leaflet and fluoresces, allowing the spermatid PMs to be visualized (Fig. 3A,B). When MOs fuse during the maturation of spermatids into spermatozoa, they leave stable membrane invaginations, which show as bright puncta around the cell body periphery in the presence of FM 1-43 (Fig. 3C,D). Up to five individual MOs can be counted accurately, thereafter the fluorescence merges and multiple MO fusions make the cell body periphery too bright to distinguish individual MOs. The formation of these puncta can be followed in real time since FM 1-43 does not interfere with spermatozoa development.

No MO fusion was observed in either triethanolamine (TEA)-activated spermatozoa from fer-1 mutants with stop codons (hc47 and eb7, n>400, each), which have no FER-1 (see following section), or in two non-conditional missense mutants (hc80 and hc136), although spermatozoa from all mutants still formed short pseudopods (Fig. 3E,F). Together with the previous observation that fer-1(hc1)/nDf23 has the same defective phenotype as hc1/hc1 (L'Hernault et al., 1988), this confirms that lack of MO fusion is the null phenotype.

To look for subtle effects that might be caused by the different fer-1 temperature-sensitive mutations in the C2 and DYSF domains, we compared MO fusion in mutant and wild-type TEA-activated spermatozoa at 15°C, 20°C, and 25°C. After 20 minutes, MO fusions in activated spermatozoa from hc1ts, hc13ts and hc24ts mutants were indistinguishable from wild-type at the permissive temperature of 15°C (Fig. 3A-D,I). The hc82ts cells had a more severe phenotype with only 6% of cells displaying normal MO fusions. At 20°C, MO fusions in both hc24ts and hc82ts mutant sperm were reduced. Since these mutations both lie in the same domain, MO fusion may be particularly sensitive to changes in this domain. At 25°C, none of the mutants had normal MOs fusions when compared to wild type (Fig. 3G,H), although 10% of hc24ts sperm underwent one or two MO fusions, consistent with its leaky phenotype at 25°C (Ward et al., 1981). These results show that although FER-1 contains six predicted C2 domains, for three of them, just a single amino acid substitution results in defective MO fusion. The mutant phenotypes, together with the sequence analysis, suggest each of the C2 domains is necessary, so their functions cannot be redundant.

To study FER-1, we developed two affinity purified rabbit polyclonal antibodies to regions indicated in Fig. 1. The anti-DysfNC antibody was used for western blotting, and the anti-peptide antibody, AZ10, which did not work for western blots, was used for immunofluorescence studies (see below). As a control for antibody specificity we used fer-1(hc47) males, which have a stop codon prior to either the anti-DysfNC or the anti-AZ10 antigenic sites, and presumably have no FER-1 protein. We compared wild-type male and fer-1(hc47) mutant male worm extracts because it was difficult to isolate enough mutant males from which to purify spermatids. On western blots, three proteins that labeled with anti-DysfNC were detected in wild-type males, each of which were missing from fer-1(hc47) males (Fig. 4A). The largest protein migrated about the expected size of FER-1 (230 kDa), but two additional bands were observed at 195 and 180 kDa in male worm extracts. Since all three bands were missing from fer-1(hc47) males, this suggested they were likely to be isoforms of FER-1. When we examined proteins from isolated spermatids, we observed the same doublet of 195 and 180 kDa proteins, which were absent from sperm-less hermaphrodites confirming their sperm specificity (Fig. 4B). Full-length FER-1 was present, but barely detectable in spermatids. This difference might be due to the presence of spermatocytes in the males, which are absent in purified spermatid preparations. If so, this suggests that the full-length FER-1 protein is present early in spermatogenesis and then becomes less abundant in spermatids, perhaps because it is proteolytically processed into the smaller isoforms.

To ensure that the 195 and 180 kDa proteins were derived from FER-1, we immunoprecipitated proteins from wild-type spermatid extracts using anti-DysfNC, separated these by SDS-PAGE and isolated the bands. These were subjected to LC-MS/MS, and peptide masses were matched against the C. elegans proteome. Insufficient protein was obtained to identify full length FER-1, but multiple peptides matching FER-1 were identified from the 195 and 180 kDa bands confirming that they were encoded by the fer-1 gene (Fig. 4C; supplementary material Table S2).

The presence of multiple FER-1 isoforms might have arisen by alternative splicing, since multiple splicing variants are observed for human dysferlin, myoferlin, and otoferlin, and some of these are known to be translated (Bashir et al., 1998; Ho et al., 2004; Liu et al., 1998; Salani et al., 2004; Yasunaga et al., 2000; Yasunaga et al., 1999). To test this we isolated additional fer-1 cDNAs from a sperm-enriched cDNA library by PCR. In addition to the full-length 6.2 kb transcript, we found only two new cDNA clones, 3.0 and 2.8 kb. These were too small to encode proteins corresponding to the 185 and 195 kDa bands (see supplementary material Fig. S3). Thus, alternative splicing is unlikely to explain these FER-1 isoforms, suggesting they might arise from proteolytic cleavage.

Since fer-1 mutants show defects in MO fusion, and FER-1 is a predicted type II (cytoplasmic facing) membrane protein, we anticipated that FER-1 would localize to MO membranes. The second antibody, anti-AZ10, was utilized for immunolocalization of FER-1, since anti-DysfNC antibodies cross reacted with an additional protein band on western blots (110 kDa) that was not eliminated in the fer-1(hc47) mutant. With anti-AZ10, punctate staining was observed around the periphery of permeabilized spermatids, confirming that sperm contain FER-1 (Fig. 5B). These puncta were eliminated in fer-1(hc47) mutant sperm (Fig. 5J).

To determine if FER-1 puncta corresponded to MOs, spermatids were co-stained with 1CB4, an antibody that labels MOs in sperm as well as the sperm PM and other structures in a few other worm cells (Okamoto and Thomson, 1985). Co-localization was only observed on the MOs, but not on the PM (Fig. 5C,D). In mature sperm, 1CB4 still labeled punctate structures only in the cell body, presumably some fused and unfused MOs, and the PM weakly (Fig. 5G). Anti-AZ10, however, only partially co-localized with these punctate structures in spermatozoa, with most of the signal observed in the PM (Fig. 5H) including distinct labeling of the pseudopod membrane. These antibody labeling experiments indicate that FER-1 is localized to MO membranes in spermatids and moves into the PM upon MO fusion during sperm activation. This pattern is similar to that previously observed for the MO-localized Ca²⁺ channel, TRP-3 (=SPE-41) (Xu and Sternberg, 2003), but differs slightly from the tetraspanin SPE-38, which is also found in the MO using an antibody to an extracellular domain (Chatterjee et al., 2005).

We utilized quantitative immunoelectron microscopy to localize FER-1 more precisely. Isolated spermatids were activated with TEA to form spermatozoa, prepared for immunoelectron microscopy, sectioned, labeled with anti-AZ10 or 1CB4 antibodies and detected with immunogold labeled secondary antibodies.

Anti-AZ10 predominately labeled MOs in spermatids, with no significant difference between labeling on the head and body (Fig. 6A,E). FER-1 localization decreased in MO body membranes upon fusion, shown by a reduction of immunogold density between unfused and fused MO membranes (Fig. 6B,G; see Materials and Methods for calculation). In parallel, the labeling on the PM increased two-fold following MO fusion, supporting the light microscope observations that FER-1 moves into the PM upon fusion. PM labeling was similar between the pseudopod and the cell body, suggesting that FER-1 is not restricted in its movement within the PM (Fig. 6G, inset). In addition, we observed an increase in localization over the pseudopod cytoplasm, which is higher than in the cell body either before or after activation (Fig. 6B,E).

Quantification of 1CB4 antibody labeling revealed its highest density on MOs in spermatids, with a slight enrichment in the MO body versus the MO head (Fig. 6C,F). In contrast to FER-1, 1CB4 immunogold density increased along MO membranes upon fusion (Fig. 6D,H). This increase occurs despite the release of MO luminal contents, indicating that the unidentified antigen recognized by 1CB4 is membrane associated. Thus, different MO proteins can have different membrane distributions following MO fusion.

Since the pattern of FER-1 membrane labeling resembled that of TRP-3 by light microscopy, we used the anti-TRP-3 antibody (gift from P. Sternberg, Caltech, CA, USA) to examine its distribution by electron microscopy. Because the total labeling for anti-TRP-3 was much lower than that for anti-AZ10 or 1CB4, we scored for the fraction of gold particles that localized to the MO head versus MO body. MO head membranes had 23% of TRP-3 labeling (n=83) whereas AZ10 and 1CB4 had 16% and 20% labeling of these compartments, respectively. Since the head membrane comprises approximately 20% of total MO membranes in these cross sections, all three antigens have a similar distribution in the head and body MO membranes. This suggests that these MO protein components are not localized to specific parts of this organelle prior to its fusion with the PM.

Many exocytotic events are stimulated by a rise in Ca²⁺ concentration. Since FER-1 has six putative C2 domains, which might bind Ca²⁺, and single missense mutations within three of these domains block MO fusion, it is reasonable to predict that Ca²⁺ triggers MO fusion. Previous experiments, however, showed that external Ca²⁺ was not the source of this activity, since MO fusion still proceeds in the presence of 1 mM EGTA (Shakes and Ward, 1989).

Using the membrane-permeable Ca²⁺ chelator BAPTA-AM, we asked if depletion of intracellular Ca²⁺ stores would prevent MO fusion. Spermatids were incubated with BAPTA-AM for 30 minutes at 15°C and triggered to fuse their MOs with TEA. MO fusion was again monitored with FM1-43. No effect was seen at 1 μM, but the fraction of cells with fused MOs dropped from 21% to 4% as the BAPTA-AM concentration increased from 2 to 10 μM (Fig. 7A,B). These results suggest that Ca²⁺ from an internal source is indeed required for MO fusion.

If MO fusion responds to changes in intracellular Ca²⁺, and if at least some FER-1 C2 domains sense Ca²⁺, we reasoned that the single amino acid substitutions found in the fer-1 C2 domain mutants might alter the sensitivity of the cell to the Ca²⁺ changes leading to MO fusion. Therefore, we tested sperm from fer-1 temperature-sensitive mutants with C2 domain mutations for MO fusion sensitivity to Ca²⁺ depletion. We did this at permissive temperature, where the protein must still function sufficiently to allow MO fusion when Ca²⁺ levels are normal. With 1 μM BAPTA-AM treatment MO fusion was normal in wild-type sperm but was reduced in the four fer-1 mutants tested (Fig. 7C,D). All mutants were significantly different from wild type (P<0.001), and hc1ts, hc24ts and b232ts were all significantly different from their EGTA-treated counterpart (P<0.01). Since MO fusion in fer-1 C2 domain mutants is more sensitive to a reduction of internal Ca²⁺ than wild type, this supports the argument that FER-1 responds to a Ca²⁺ signal, and further suggests that the C2 domains are involved in this Ca²⁺ response.

fer-1 was originally identified in sterile mutants with defective sperm. Two alleles were found to be defective in the fusion of sperm MOs with the PM. When the fer-1 gene was identified, its sequence gave little insight into its function. Subsequently, homologs have been found in many animals, including two proteins associated with human diseases, and this family of proteins was named `ferlin' after its founding member. Here, we have focused on analyzing the membrane fusion phenotype in order to elucidate the function of FER-1.

Using a simple light microscopic assay to detect MO fusions, we found that all of the fer-1 mutations cause defects in MO fusion. This includes seven missense mutations, which fall within predicted functional domains, one missense mutation in a conserved region between domains, and two mutants that generate stop codons. Since both western blotting and immunofluorescent staining show FER-1 protein(s) are eliminated in the stop codon mutant fer-1(hc47), the MO fusion phenotype must be the null phenotype, as predicted from previous genetic data (L'Hernault et al., 1988). In addition, the only function of FER-1 appears to be its role in sperm MO fusion since fer-1 mutants have no phenotype other than sterility and expression of fer-1 is only detected during spermatogenesis (Achanzar and Ward, 1997; Reinke et al., 2000).

Using an antibody specific for FER-1, we found that FER-1 is localized to MOs prior to fusion, and some of it moves into the PM following fusion. Detailed analysis of FER-1 localization by immunoelectron microscopy determined that FER-1 is not confined to specific membranes within the MO structure, but has a similar density in the head and body membranes of the MO. This is also true for the MO marker 1CB4 antigen and for the previously studied Ca²⁺ channel TRP-3 (Xu and Sternberg, 2003). However, after fusion the distribution of the 1CB4 antigen differs from that observed for FER-1 and TRP-3. It does not move past the electron dense collar, resulting in a net increase of 1CB4 antigen along membranes remaining in the fused MO cavity. These results show that various MO-localized proteins can have different distributions following fusion, implying that the MO itself or the fusion process can regulate membrane protein distributions.

Although the fer-1 mutants were obtained from random mutagenesis screens, the resulting mutations are non-randomly distributed throughout the fer-1 sequence, with all but one located in predicted functional domains. Two mutations lie within the DysfNC domain, a nested repeat sequence in many ferlins and several other proteins, whose function is unknown. Missense mutations within this domain in dysferlin have been linked to Miyoshi myopathy (Aoki et al., 2001; Matsumura et al., 1999) also indicating its importance.

Five fer-1 mutations (hc1ts, hc136, b232ts, hc82ts, hc24ts) fall within three of the predicted C2 domains. This suggests that these domains are essential for FER-1 function and are particularly sensitive to amino acid substitutions. Since single missense mutations in these C2 domains result in defective MO fusion, each domain must have distinct, non-overlapping functions. This conclusion is further supported by sequence analysis of the C2 domains in ferlins from several species, including human dysferlin. Phylogenetic analysis of ferlin C2 domains reveals that C2 domains at a similar position in different ferlins are more similar to each other than to the other C2 domains within the same protein. Since most ferlins have retained the full complement of C2 domains over evolution, this suggests that each must have an important function, and that the positional order must also be important to overall ferlin function, most probably to ensure their orientation in the folded protein.

Vesicle fusion in a wide variety of cell types is stimulated by a Ca²⁺ increase (Bi et al., 1995; Borgonovo et al., 2002; Eddleman et al., 1997; Reddy et al., 2001; Savina et al., 2003; Steinhardt et al., 1994) and detected by the two C2 domains of the protein synaptotagmin, located on the vesicle surface (Bai and Chapman, 2004; Brose et al., 1992). The presence of multiple C2 domains in FER-1, as well as its location on the MO membrane, suggests that it might be directly involved in responding to a Ca²⁺ signal. Although extracellular Ca²⁺ is not required for MO fusion, we found that the intracellular chelator BAPTA-AM could block MO fusion in a dose-sensitive manner when applied to spermatids prior to their activation to form spermatozoa. Thus, like most other membrane fusion processes, Ca²⁺ appears to be essential for MO fusion.

Further evidence that FER-1 is directly involved in responding to a Ca²⁺ signal is that fer-1 mutants with temperature-sensitive missense mutations display hypersensitivity to a reduction in free Ca²⁺ with BAPTA-AM, even at the permissive temperature of 15°C. Whereas wild-type sperm activated normally at 15°C, fer-1ts mutant sperm formed pseudopods but failed to fuse MOs, similar to the fer-1 null phenotype. This was observed in all four of the C2 domain mutants tested. Although it is possible that these C2 domain missense mutations might act indirectly by altering FER-1 structure, because the sperm from each mutant showed a MO fusion defect when free calcium was reduced at the permissive temperature, this result provides strong evidence that the C2 domains of FER-1 directly mediate the FER-1 response to a Ca²⁺ signal to trigger MO fusion.

Vesicle fusion during the repair of human muscle membranes requires the fer-1 homolog dysferlin, as well as a localized spike in Ca²⁺ concentration at the sites of fusion, suggesting a role for Ca²⁺ in this ferlin-mediated fusion process (Bansal et al., 2003). In C. elegans sperm, a Ca²⁺ trigger could promote phospholipid binding or interactions with other proteins directly involved in membrane fusion, such as SNAREs. In vitro experiments have shown that the N-terminal C2 domain of dysferlin and myoferlin (C2A) is capable of Ca²⁺-dependent phospholipid binding, which is disrupted when a disease-causing mutation is introduced (Davis et al., 2002). Our phylogenetic analysis also shows that the C2A domain is the only ferlin C2 domain that clusters together with the majority of characterized C2 domains, which may explain its unique biochemical behavior. Because fer-1 does not have this homologous C2 domain, yet is sensitive to Ca²⁺ depletion, it suggests that a Ca²⁺ signal does more than promote phospholipid binding by the remaining C2 domains. For example, the C2 domains together might act like a scaffold to modulate protein-protein interactions, perhaps with SNARE proteins or other membrane fusion components. The conserved order of the different C2 domains might specify a spatial organization necessary for these interacting proteins.

We found that there were three FER-1 protein isoforms in sperm that were missing in the fer-1(hc47) mutant. The smaller proteins were confirmed by mass spectrometry to be fragments of FER-1. One explanation for the smaller protein bands is that the variants could arise by alternative splicing or alternate start sites. It has been shown that the human homolog OTOF has long (240 kDa) and short (130 kDa) isoforms, which are generated with alternative transcriptional start sites, and multiple smaller dysferlin transcripts have been identified in various tissues. We isolated two new fer-1 cDNAs from a sperm-enriched library and by RT-PCR, but they were too small to encode the observed FER-1 isoforms. Thus, we found no evidence that the smaller FER-1 proteins arise from alternative splicing, although we cannot rule out alternative start sites.

The relative proportion of FER-1 isoforms was different between protein samples prepared from whole males versus purified spermatid preparations. The 230 kDa (full length) FER-1 was more abundant in males, and reduced in spermatids, whereas the 195 and 180 kDa isoforms were abundant in both. Unlike males, purified spermatid preparations include few spermatocytes, so it is likely that the full-length FER-1 protein is made in spermatocytes and then cleaved to form the smaller isoforms during the maturation of spermatocytes into spermatids. No matter the source of the smaller isoforms, their peptides found by mass spectrometry include at least the domains between C2C and C2E and from their size must include additional domains, but the exact ends of these isoforms are not known. We do not know the function of this cleavage, but a possible candidate to cleave FER-1 is the protein SPE-4, which is also localized to the MO. SPE-4 is a divergent member of the presenilin family, which is involved in intramembranous proteolysis leading to Alzheimer's disease (Arduengo et al., 1998; Brunkan and Goate, 2005). It is a surprising coincidence that two proteins related to human diseases are both found in such an unusual organelle as the nematode sperm MO, so this may suggest some connection between these human diseases.

Based on our results, we propose a model whereby FER-1 regulates Ca²⁺-dependent vesicle fusion as depicted in Fig. 8. In spermatids, MOs are `docked' with their heads abutting the PM as the cells await the activation signal. This localization of the MO is not fer-1 dependent, since MOs appear to be docked appropriately in fer-1 mutants (Ward et al., 1981). FER-1 protein is localized uniformly to the MO membrane surface, but because of its orientation, or other as yet unidentified head-specific proteins, fusion occurs between the head membrane and PM. Given an activation signal, spermatozoa utilize intracellular Ca<sup>2+</sup> to relay this signal, with FER-1 acting to relay a `fusion' signal, which results in fused MOs at the point of contact between their head and the PM. As a result, FER-1 and some other MO membrane proteins mix into the PM, while other membrane proteins remain trapped behind the fusion collar.

This model for FER-1 is similar to the model proposed by Bansal and Campbell (Bansal and Campbell, 2004) for the role of dysferlin in membrane repair. Both proteins reside in cytoplasmic vesicles that fuse with the PM in response to a Ca²⁺ signal. Once the vesicles fuse, FER-1/dysferlin is present in the PM. fer-1 mutants and their Ca²⁺ sensitivity demonstrate the importance and non-redundancy of the multiple C2 domains in the ferlin family. Taking advantage of the C. elegans genetic system, future studies to determine the exact function of FER-1 in MO fusion might identify additional functions that are shared with dysferlin and other ferlin family members.

Culture, manipulation of worms and genetic analyses were performed with standard methods (Brenner, 1974). All strains used in this work were derived from the wild-type C. elegans strain var. Bristol N2. Standard C. elegans nomenclature has been used throughout this paper (Horvitz et al., 1979). The isolation of fer-1 (hc1ts, hc13ts, hc24ts, hc80, hc91ts, b232ts, hc47, hc82ts) was described previously (Argon and Ward, 1980; L'Hernault et al., 1988). The ethyl methane sulfonate-induced allele fer-1(eb7) was a gift from S. L'Hernault (Emory University, Atlanta, GA). Genetic markers used were LGI, dpy-5(e61); LGIV, fem-1(hc17ts) (Nelson et al., 1978) and fem-3(q20ts) (Barton et al., 1987); LGV, him-5(e1490) (Hodgkin et al., 1979). Non-conditional alleles hc47, eb7, hc80 and hc136 were marked by linkage to dpy-5, and were maintained by picking Dpy L4 hermaphrodites, checking for sterility, and mating back to wild type. All strains were maintained at 15°C, unless otherwise stated. The gene T05E8.1 was named ferl-1 (fer-like) based on its sequence similarity to fer-1.

Genomic DNA from temperature-sensitive mutants was obtained by washing off mixed-stage populations with buffer and adding PCR lysis buffer (50 mM KCl, 10 mM Tris-HCl, pH 8.2, 2.5 mM MgCl₂, 0.45% NP-40, 0.45% Tween 20, 0.1 mg/ml gelatin). For hc47 and eb7 non-conditional strains, dumpy hermaphrodite progeny of fer-1 dpy-5/++ were transferred to individual 6 mm plates and scored 24 hours later for self-sterility. Sterile worms were picked directly into a 5 μl drop of PCR lysis buffer. Worms were frozen at -20°C and a worm lysate was prepared (Barstead and Waterston, 1991). For DNA sequencing, 20 overlapping fer-1 gene segments were amplified from worm lysate by PCR with fer-1-specific primers, in quadruplicate, using Taq DNA polymerase (Promega Corp., Madison, WI, USA). Mutations were identified by sequencing PCR fragments in both directions (Genomic Analysis and Technology Core Sequencing Facility, The University of Arizona, USA). Only one mutation was found in each strain.

Sequences were obtained from Ensembl (www.ensembl.org) and analyzed with VectorNTI Advance 9.0. A table of accession numbers for ferlin protein sequences is provided in supplementary material Table S1. Approximate C2 domain boundaries were initially determined by SMART (smart.embl-heidelberg.org) and adjusted by eye. Ferlin C2 domain sequences were aligned to an existing alignment of non-ferlin C2 domains generated by Nalfeski and Falke (Nalfeski and Falke, 1996) using ClustalW with default parameters, followed by adjustment by eye. A tree search was conducted using neighbor-joining algorithm in PAUP^* 4.0 (Swofford, 2003). Bootstrap values were obtained with 1000 replicates, with 10 random sequence additions per replicate. The tree represents majority-rule consensus.

For antisera preparation one series of rabbits was immunized with a partial FER-1 fusion protein representing amino acids 723-957, consisting of the DysfN-DysfC domains fused in-frame after Schistosoma japonicum glutathione S-transferase (GST) in the pGEXM7 bacterial expression vector (gift from David Drechsel, Max Planck Institute for Cell Biology and Genetics, Dresden, Germany). Soluble protein was purified from E. coli extracts using a glutathione-Sepharose fast flow affinity column according to the manufacturer's instructions (Amersham Biosciences). Purified GST-DysfNC protein was used to inoculate rabbits and serum from these was selected by ELISA assay (Eurogentec, Herstal, Belgium). The same fusion protein was covalently coupled to CNBr-activated Sepharose (Amersham) to create an affinity column. Anti-DysfNC antisera were first cleared of any anti-GST and anti-E. coli antibodies by passage through GST and E. coli affinity columns, followed by affinity purification over the GST-DysfNC column following the manufacturer's instructions. Purified anti-DysfNC antibodies were combined and concentrated for a final stock concentration of 2 mg/ml.

A second set of rabbits was immunized (Cocalico Biologicals, Reamstown, PA) with the synthetic peptide CKSMKGDFDDPEEKEK corresponding to residues 1334-1348 with an N-terminal cysteine residue [Macromolecular Structures Facility, Arizona Research Laboratories (W. E. Achanzar, Analysis of a gene required for membrane fusion during nematode spermiogenesis. PhD Thesis, The University of Arizona, 1996.)]. Antiserum (anti-AZ10) was purified by peptide affinity chromatography and concentrated as above. Final stock concentration of resulting anti-AZ10 peptide antiserum was 4 mg/ml.

Spermatids were isolated from fem-3(q20) hermaphrodites or him-5(e1490) males as previously described (Nelson et al., 1982). For western analysis using handpicked males, individual L4-virgin him-5 or fer-1(hc47) dumpy male worms were picked onto plates and allowed to accumulate sperm without mating for 2-3 days at 20°C. These were then picked and frozen at -20°C. Worms and/or sperm protein preparations for SDS-PAGE were resuspended in sperm medium, pH 7.8 [SM; Nelson and Ward (Nelson and Ward, 1980)], + 10 mg/ml glucose [SMG; Machaca et al. (Machaca et al., 1996)], sonicated, and TCA precipitated. Protein pellets were taken up in 1 × sample buffer, boiled, and electrophoresed on 15×20 cm 5-20% (0.4:1 acrylamide:PDA crosslinker) gradient gels and blotted onto Hybond-P membranes. Blots were probed with anti-DYSFNC 1:10,000, followed by HRP-conjugated donkey anti-rabbit (Pierce) at 1:50,000, and visualized with ECL-Plus Chemiluminescent Reagent (Amersham). The anti-AZ10 antibody does not work on western blots, although it works properly by indirect immunofluorescence on cells.

For immunoprecipitation from spermatids, 1.5×10⁸ frozen isolated cells were homogenized by sonication in 500 μl IP buffer (300 mM NaCl, 10 mM NaHPO₄ pH 7.5, 5 mM EGTA, 0.2 mM EDTA, 1 mM MgCl₂, 0.02% azide) with protease inhibitors (Complete, Roche) [adapted from Fowler et al. (Fowler et al., 1993) and Gregorio and Fowler (Gregorio and Fowler, 1995)]. To lysates, an equal volume of boiling IP buffer containing 0.4% (w/v) SDS was added, and samples boiled for 2 minutes. After cooling to room temperature, 100 μl/ml 20% Triton X-100 was added and gently vortexed. Insoluble material was sedimented at 66,000 g for 1 hour at 4°C, and soluble extracts were transferred to prepared beads.

Washed protein A-Tris-acryl beads (20 μl, packed; Pierce) were incubated with 50 μg antibody for 1 hour with constant mixing. Immunoprecipitations were carried out overnight at 4°C with end-over-end rotation. In parallel, control beads without bound antibodies were also incubated with clarified spermatid extract. After binding, beads were washed four times with IP buffer (without SDS). Washed beads were resuspended in 2× SDS sample buffer, boiled, and separated by SDS-PAGE (Laemmli, 1970). For mass spectrometry, gels were stained with silver and protein bands were excised, trypsin digested and eluted [adapted from Blum et al. (Blum et al., 1987) and Shevchenko et al. (Shevchenko et al., 1996)]. Peptides were identified by liquid chromatography-mass spectrometry/mass spectrometry (LC-MS/MS) by L. Breci at The University of Arizona Proteomics Facility. Peptide masses were searched using TANDEM (http://www.thegpm.org), with WormPep release WS140 (May 2005).

Sperm from him-5 males were dissected into 10 μl SMG on a slide coated with 1 mg/ml poly-L-lysine (Sigma), fixed, and permeabilized as described previously (Nelson and Ward, 1980). For activation, selected slides were treated with 100 mM triethanolamine (TEA, Sigma), pH 7.8 for 15 minutes. Nonspecific binding was blocked with BSA blocking solution (1% BSA + 0.01% Tween 20 in PBS) for 1 hour. Primary antibodies diluted in this blocking solution were added to slides for 1 hour at room temperature (or 4°C overnight), followed by washing and then secondary antibodies were added for 30 minutes. 4,6-diamidino-2-phenylindole (DAPI, 1 mg/ml in BSA-block) was added for 15 minutes, and quickly rinsed off. Samples were mounted in Prolong Antifade mounting medium (Molecular Probes).

Affinity purified anti-AZ10 peptide antibodies were used at a dilution of 1:100 for immunofluorescence. The cell line that secretes the monoclonal antibody 1CB4 (Okamoto and Thomson, 1985) was provided by J. Ahringer and J. Hodgkin, and culture supernatant was prepared by C. Heilman and A. I. Levey (Department of Neurology, Emory University School of Medicine, Atlanta, GA). 1CB4 hybridoma culture supernatant was used for immunofluorescence at 1:5,000.

Mouse monoclonal primary antibodies were visualized by using Alexa Fluor 647-conjugated goat anti-mouse IgG secondary antibodies (Molecular Probes). Rabbit polyclonal antibodies were detected with Alexa Fluor 488-conjugated goat anti-rabbit secondary antibody (Molecular Probes). All microscopy employed a Leica DMRXA microscope fitted with a 100× Plan Apo objective and 1.6× Optivar. Micrographs were taken through appropriate filters (Cy5 for Alexa Fluor 647, FITC for Alexa Fluor 488, and DAPI; Chroma Technology Corp., Brattleboro, VT) with a Retiga EX digital camera. Z-series stacks of cell images were captured, deconvolved, and reconstructed with MetaMorph 6.2 software.

Spermatids were isolated in bulk and activated with 100 mM TEA for 10 minutes. Cells were fixed in 4% paraformaldehyde + 0.2% gluteraldehyde in SM, pH 7.8, and embedded in LR White. Blocks were prepared by the Electron Microscopy Facility at The Max Planck Institute for Cell Biology and Genetics (Dresden, Germany). Ultrathin sections were mounted on nickel-coated grids. All incubations were performed at room temperature in a humid chamber, with grids submerged in 10 μl drops of medium on Parafilm. Grids were blocked on both sides with 0.5% BSA in PBS for 30 minutes, followed by incubation with AZ10 (1:10) or 1CB4 (1:500) antibodies for 1 hour. Washing was performed by transfer of grids through three 0.1% BSA + 0.05% Tween 20 in PBS drops, 5 minutes each. Primary antibodies were detected with 10 nm immunogold goat anti-rabbit or goat anti-mouse secondary antibodies (diluted 1:20; Electron Microscopy Sciences, Hatfield, PA), incubated for 1 hour, followed by washing as above. Antibody-labeled grids were fixed with 4% paraformaldehyde + 0.2% gluteraldehyde in SM for 2.5 minutes, rinsed in water, stained with saturated uranyl acetate (10 minutes), rinsed with water and dried. Images were captured on a Phillips 420 transmission electron microscope with Kodak film. Negatives were digitized, and distance/area measurements were acquired with Metamorph software.

Equation 1 was then used to estimate the corresponding membrane length for unfused MOs in immunogold-labeled samples, since the area of the MO was readily measured. Similarly, we used equation 1 to determine the length of membrane remaining invaginated within the MO cavity following fusion, given the area occupied by electron-dense material. The results obtained from these membrane measurements were similar to those reported previously based on line-intercept morphometry (Roberts et al., 1986).

For temperature-sensitivity assays, picked him-5 hermaphrodites were allowed to lay eggs at 15°C, 20°C, or 25°C. Resulting L4 virgin males were picked onto separate plates to prevent mating. Spermatids were dissected from these males onto poly-L-lysine-coated slides into 50 μl Ca²⁺-free SMG + 10 mM EGTA, pH 7.8 (SMGE). For BAPTA-AM experiments, spermatids were incubated in SMGE containing 0.001% (v/v) Pluronic-127 with or without BAPTA-AM at varying concentrations for 30 minutes then washed for 10 minutes at 15°C in SMGE. A flow chamber was created by raising the coverslip with a thin application of Vaseline on 2 parallel sides (Shakes and Ward, 1989). Cells were mounted in SMGE with 2 μM N-(3-triethylammonium-propyl)-4-[4-(dibutylamino)styryl]pyridinium dibromide (FM1-43; Molecular Probes) to visualize membranes. Spermatids were activated at room temperature by flowing 100 mM TEA in SMGE with FM1-43 through the chamber. DIC and FITC filtered images were captured every 20 seconds for 20 minutes. Longer times did not lead to additional fusions. Surrounding fields of cells were collected and counted. Up to five individual MO fusions could be scored easily, thereafter the MO fluorescence merged so that individual MOs could not be distinguished. Spermatozoa with ⩾5 fused MOs were scored as `normal', although wild-type cells have an average of 18-25 MOs, and approximately 70% of them fuse at 25°C (40% fuse at 16°C) (Ward et al., 1981). Lysed cells, which were identified by the entire contents of the cell fluorescing, were not counted.

We thank Sarah McCarthy for help with DNA sequencing, David Dreschel, Bill Achanzar and Michael Galligan for antibody preparation, and David Bentley for help with electron microscopy. We also thank Asher Cutter and Matthew Terry for help with statistics and phylogenetic analysis. We thank Carol Dieckman, Johnny Fares, Carol Gregorio and Lisa Nagy for advice, members of the Fares and Nagy laboratories for helpful discussions, and an anonymous reviewer for comments on the text. N.W. was supported by NIH training grants T32-CA09213 and T32-A09213. This work was supported in part by NIH grant R01 GM 25243.