ABSTRACT
During maturation of spermatids to motile spermatozoa in Caenorhabditis elegans, large vesicles called membranous organelles (MOs) fuse with the spermatid plasma membrane. Mutations in the gene fer-1 cause abnormal spermatozoa in which the MOs do not fuse, although they abut the plasma membrane normally. Here we describe the fer-1 gene, which we found to be approximately 8.6 kb in length and to encode a 6.2 kb transcript whose expression is limited to the primary spermatocytes, the cells in which the MOs form. fer-1 is predicted to encode a 235 kDa protein which is highly charged except for a putative transmembrane domain near the C terminus. We identified the mutations associated with five fer-1 alleles, all of which are missense mutations causing single amino acid changes. FER-1 is not similar to any characterized proteins in sequence databases, nor does it contain known functional motifs other than the predicted transmembrane domain. The C-terminal transmembrane domain makes FER-1 resemble some viral fusion proteins, suggesting it may play a direct role in MO-plasma membrane fusion. FER-1 does show significant sequence similarity to several predicted human proteins of unknown function. Two of the identified fer-1 mutations are located in regions of similarity between FER-1 and two of these predicted proteins. This strengthens the biological significance of these similarities and suggests these regions of similarity represent functionally important domains of FER-1 and the human proteins.
INTRODUCTION
Spermatogenesis in C. elegans (extensively reviewed by Wolf et al., 1978; Kimble and Ward, 1988; L’Hernault and Roberts, 1995; L’Hernault, 1997) has been used to study a number of important cellular phenomena including cell motility, cytoskeletal organization, signal transduction, aging, and sperm competition. This system offers many attractive features, the most important being its amenability to genetic analysis. Over 60 genes have now been identified that disrupt spermatogenesis when mutated, thus enabling genetic dissection of this process (L’Hernault et al., 1988, 1993; L’Hernault, 1977; Minniti et al., 1996; Varkey et al., 1995; L’Hernault and Arduegno, 1992). Focusing on individual steps of spermatogenesis has allowed examination of cellular events important not only for nematode sperm development, but also of general importance to cell biology.
Maturation of haploid spermatids (spermiogenesis) in the free-living nematode Caenorhabditis elegans involves fusion of large vesicles called membranous organelles (MOs) with the plasma membrane (Nelson and Ward, 1980). The MOs are bi-lobed vesicles consisting of a smaller head lobe and larger body lobe separated by an electron dense collar (Fig. 1). The two lobes are formed from a single contiguous membrane but differ in their contents and membrane composition based on immunological data (Roberts et al., 1986). MO-plasma membrane fusion is incomplete; only the head lobe fuses and the body forms an invagination on the cell surface connected to the exterior by the fusion pore as shown in Fig. 1 (Nelson and Ward, 1980). The MO contents are mainly glycoproteins based on protease sensitivity and lectin staining (Ward et al., 1981). This glycoprotein is released during fusion and is seen in transmission electron micrographs as fibrous material extending from within the invaginations, through the fusion pore, and around the cell surface after fusion. The MOs also contribute membrane and glycoproteins to the plasma membrane (Roberts et al., 1986).
The precise function of the MOs and their contents during spermiogenesis is not known, but MO fusion is essential since mutations in the gene fer-1 block fusion and cause defective, infertile spermatozoa. Ten recessive fer-1 alleles have been isolated, four are non-conditional and six are temperature sensitive (Ward and Miwa, 1978; Argon and Ward, 1980; L’Hernault et al., 1988; see Table 1). They all have a similar sperm-defective phenotype: no fusion of the MOs with the plasma membrane, formation of short pseudopods and inability to crawl although their pseudopod membrane is motile (Argon and Ward, 1980; Ward and Miwa, 1978; Ward et al., 1981). This is likely to be the null phenotype since it is obtained for several of the mutations over a deficiency (L’Hernault et al., 1988). The mutant sperm also have defects in surface membrane flow (Roberts and Ward, 1982). The MOs appear normal in fer-1 mutant sperm and orient properly with their head lobes up against the plasma membrane as they would in a normal spermatid (Fig. 1). This suggests the fer-1 gene product is not involved in MO localization but is required for fusion with the plasma membrane. We present here the isolation and molecular characterization of the fer-1 gene and its product.
MATERIALS AND METHODS
Worm strain handling
C. elegans var. Bristol, strain N2, was used as the wild-type. Sperm was often obtained from a strain carrying the mutation him-5(e1490), which increases the frequency of males in the population but has no known effect on the sperm (Hodgkin et al., 1979). All strains were maintained on agar plates seeded with Escherichia coli strain OP50 and manipulated as described by Brenner (1974). The fer-1 mutant strains used and their temperature sensitivities are summarized in Table 1.
DNA transformation by microinjection
All DNA transformations were done using BA576: fer-1(hc13ts)I; him-5(e1490)V hermaphrodites. hc13ts is a temperature sensitive mutant that is completely sterile at 25°C, but has wild-type fertility at 16°C and 20°C (Argon and Ward, 1980; L’Hernault et al., 1988). Cosmids (obtained from A. Coulson, Sanger Centre, Cambridge, UK) and plasmids were isolated using Qiagen columns following the manufacturer’s protocols (Qiagen, Inc.) and were resuspended in filtered TE (10 mM Tris-HCl, 1 mM EDTA), pH 7.6. Restriction digested DNA was phenol-chloroform extracted, ethanol precipitated, and resuspended in filtered TE, pH 7.6.
The DNA was microinjected into one or both gonadal arms of homozygous mutant hermaphrodites raised at 16°C (so they would be fertile) following published procedures (Mello et al., 1991). Test DNAs were injected either in the presence or absence of the dominant marker plasmid pRF4 (kindly provided by C. Mello). Worms carrying the marker plasmid were easily recognized since they rolled while crawling. Injected DNA concentration was 100-400 ng/ml in filtered TE, pH 7.6. When used, the marker plasmid concentration was 50100 ng/ml. After injection, worms were transferred to agar plates seeded with OP-50, and their F1 or F2 progeny were scored for rescue of sterility at the restrictive temperature of 25°C.
cDNA library screening
A C. elegans UniZap library (Stratagene) made from RNA extracted from male-enriched populations was plated out, plaques were lifted onto Magnagraph nylon filters (MSI), and the DNA was UV crosslinked to the filters using a Stratalinker (Stratagene) as described by Querter-mous (1996). Filters were hybridized with a digoxygenin-labeled probe made from a 2.1 kb PacI-SacI fragment from the fer-1 genomic clone. Probe synthesis, hybridization, and detection were performed using the Genius system (Boehringer Mannheim). cDNAs were excised from the phage by incubation with VSV helper phage and subsequent transformation into E. coli XL-1 Blue cells (Stratagene). The resulting plasmids contained the cDNAs in the pBluescript SK- vector (Stratagene).
Differential northern blots
Total RNA was isolated using Trizol (Gibco-BRL) from him-5(e1490) males, fem-1(hc17ts) hermaphrodites raised at 25°C (produce oocytes but no sperm, Nelson et al., 1978; Doniach and Hodgkin, 1984), and fem-3(q23ts) hermaphrodites raised at 25°C (produce sperm but no oocytes, Barton et al., 1987). A 30 μg sample of total RNA from each strain was electrophoresed for 18 hours at 40 V on a 20 cm × 25 cm agarose gel containing 1.4% formaldehyde. The RNAs were transferred to a positively-charged nylon filter (Boehringer Mannheim) by capillary transfer and UV crosslinked to the filter as above (Brown, 1996). After prehybridization, the blots were simultaneously hybridized with digoxigenin-labeled probes to fer-1 and actin (10 ng of each probe/ml) synthesized using the Genius system (Boehringer Mannheim). The template for the fer-1 probe was a 1.1 kb BamHI-EcoRI fragment from the cDNA pCDWA105 and for the actin probe a 1.1 kb C. elegans actin cDNA (kindly provided by J. Waddle) excised from the vector using EcoRI. The restriction fragments were purified from preparatory agarose gels using Qiaex resin (Qiagen) and 100 ng of each fragment was used for probe synthesis.
In situ hybridizations
In situ hybridizations were performed using the protocol of Seydoux and Fire (1995). N2 males were dissected in phosphate-buffered saline (PBS) on poly-lysine-coated slides to expose the testes. Samples were dehydrated, permeabilized, and fixed according to the published protocol. After prehybridization, single-stranded digoxigenin-labeled antisense or sense fer-1 probes were hybridized to the samples overnight at 48°C. Probes were synthesized as described using the fer-1 cDNA clone pCDWA105 as a template. The probes were detected colorimetrically by incubation with an alkaline phosphatase-conjugated anti-digoxigenin antibody (diluted 1:2,500 in PBS + 0.1% BSA, 0.1% Triton X-100) for 2 hours at room temperature, then with 4.5 μl/ml NBT and 3.5 μl/ml X-phosphate (Boehringer-Mannheim) in staining solution until color development was satisfactory. Nuclei were counterstained with the fluorescent dye DAPI added to the staining solution (1 μg/ml). The slides were examined on a Zeiss Universal microscope equipped for Nomarski differential interference contrast and fluorescence microscopy.
Extending cDNA clones with PCR
Two rounds of 5’ rapid amplification of cDNA ends (RACE) PCR were performed using the 5’ RACE kit from Gibco-BRL. Briefly, a fer-1-specific primer was used for reverse transcription using 1 μg of total RNA from fem-3(q23) hermaphrodites as template. The first strand cDNA was oligo(d)C tailed using terminal transferase and PCR performed using a nested fer-1-specific primer and an anchor primer provided with the kit. PCR products were digested with SpeI and KpnI
(1st round) or SpeI only (2nd round), then ligated into pBluescript II SK+ (Stratagene) cut with the same enzymes. The ligated DNAs were transformed into XL-1 Blue MRF’ cells and cells were plated on LB agar containing 100 μg/ml ampicillin, X-gal, and IPTG to allow for selection of recombinant plasmids by blue-white selection. Transformants were screened by colony lift and PCR for those with appropriate size RACE products.
DNA sequencing
The fer-1 cDNA sequence was determined using the dideoxy method and the Sequenase v2.0 enzyme (USB-Amersham). Primers to the pBluescript polylinker were used to sequence the ends of all cDNAs. Internal sequence was obtained by primer walking or by making nested deletions with exonuclease III (Tabor, 1996) and using vector primers. Sequencing templates were prepared in two ways: boiling minipreps (Engebrecht et al., 1996) for nested deletion products and Qiagen ‘midipreps’ for intact clones (Qiagen, Inc.).
fer-1 genomic sequencing was done at the Macromolecular Structure Facility of the Arizona Research Laboratories, Division of Biotechnology at the University of Arizona using an ABI373 automated sequencer. DNA templates were sequenced at 0.2 mg/ml using fer-1 specific primers at 5 μM. Sequencing data were analyzed and edited with the MacVector and Assemblylign sequence analysis programs (Kodak). Database searches were performed using the BLAST search tool of databases at the National Center for Biotechnology Information (Altschul et al., 1990).
The genomic and cDNA sequences were submitted to GenBank in May 1996, accession #U57652. Subsequently the region of the genome containing fer-1 has been sequenced by the C. elegans genome project. They sequenced cosmid F43G9 which overlaps C02D7, the cosmid which we found contains fer-1 as shown below. The F43G9 cosmid sequence was submitted in November 1996, accession #Z79755. The fer-1 gene is F43G9.6. The fer-1 sequences agree exactly in all the coding regions and have minor disagreements in one intron and in non-coding regions beyond the 3’ end of the gene. All references to sequence positions are to the sequence numbering in accession U57652.
fer-1 mutation identification
Genomic DNA corresponding to fer-1 was directly amplified by PCR from wild-type and homozygous fer-1 mutant worms using the protocol of Williams et al. (1992). Several worms were used for each starting DNA and several independent PCR reactions were pooled after selection of the appropriate size band on agarose gel electrophoresis. The pooled PCR products were sequenced directly using the Sequenase PCR product sequencing kit (USB-Amersham). Sequencing reactions were fractionated on 5% acrylamide gels containing 7% urea with the termination reactions for each nucleotide run side by side to help detect single base changes.
RESULTS
fer-1 cloning
fer-1 had been genetically mapped to chromosome I between lin-10 and unc-29 (Fig. 2) (Argon and Ward, 1980). Since both of these genes had been cloned and positioned on the physical map, we knew that fer-1 must be on a cosmid in the intervening region. We began microinjecting fer-1(hc13ts) worms with the cosmids indicated on Fig. 2 and found that C02D7 repro-ducibly restored partial fertility to mutant worms at the restrictive temperature (25°C). No rescue was obtained with other cosmids tested, including two that overlap C02D7 on one end. The number of progeny produced by transformed worms was highly variable (Table 2), with a mean of 82±42 per worm. Such variability is not uncommon in transformed worms and probably results from different levels of transgene expression in the gonads of individual worms (Mello et al., 1991).
In order to locate fer-1 more precisely, we injected restriction digests of C02D7 to identify enzymes that did not cut within fer-1. Fragments from rescuing digests were subcloned into pBlue-script II SK+ and injected individually to identify the one containing fer-1. Using this strategy, fer-1 was localized first to a 16 kb SpeI fragment, pWA102, then to a 10.5 kb SpeI-Sac I fragment, pWA111 (Fig. 2). As shown in Table 2, there was little difference in rescue between the intact cosmid and the subfragments, indicating that the rescuing gene appeared to be complete. Attempts to further subclone pWA111 were unsuccessful, suggesting fer-1 covered most, if not all, of this fragment.
fer-1 encodes a 6.2 kb sperm-specific transcript
The transformation experiments with the C02D7 and pWA102 restriction digests indicated the enzymes Pac I and Sac I cut within fer-1, so a 2.1 kb Pac I-Sac I probe (Fig. 2) was used to screen our cDNA library. Out of 500,000 plaques screened, we isolated a single 2.4 kb cDNA clone (pCDWA105). Sequencing suggested this represented the 3’ end of the fer-1 message since it had a 3’ poly(A)+ tail. A fragment of this cDNA was then used to probe a differential northern blot to determine the size of the corresponding transcript and to determine if it was expressed sperm-specifically. Fig. 3 shows that the probe recognized a 6.2 kb transcript present only in RNA isolated from worms producing sperm. Sperm-specific expression of fer-1 was expected since production of abnormal sperm is the only defect seen in fer-1 mutants.
The size of the transcript indicated the cDNA represented about one-third of the total messenger RNA. Attempts to isolate longer cDNAs from our library and other libraries by both plaque hybridization and PCR were unsuccessful, so we used 5’ rapid amplification of cDNA ends (RACE) PCR to isolate the remainder of the cDNA. Starting with 1 μg of total RNA from fem-3(q23) worms, we were able to amplify, clone, and sequence two cDNAs (pCDWA128 and pCDWA129) that extended an additional 3.8 kb upstream from the 5’ end of the original clone, giving a total cDNA length of 6.2 kb. This size agrees with the transcript size estimated from the northern blot.
fer-1 expression is limited in males to the testis
The cDNA pCDWA105 was used as a template to synthesize a fer-1 antisense digoxigenin-labeled probe for in situ hybridization to see where in wild-type males fer-1 is expressed. The only tissue in which fer-1 message was detected was the testis, beginning in the loop and proceeding proximally (Fig. 4A). Based on DAPI-staining of nuclei to identify the stages of meiosis, expression begins when the cells enter meiosis and is confined to the primary spermatocytes (Fig. 4C,E). No signal was detected in the secondary spermatocytes, the mitotic sper-matogonial cells, the spermatids or in any other tissue in the worm. No signal was detected using a fer-1 sense probe as a negative control (Fig. 4B). This result is in agreement with the genetic and northern analyses.
fer-1 genomic organization
Comparison of the genomic and cDNA sequences showed fer-1 to be approximately 8.6 kb in length and composed of 21 exons and 20 introns (Fig. 5). Typical of most C. elegans genes examined, the introns are small (Krause, 1995). The cDNAs almost certainly represent the entire coding region based on two criteria. The 5’ end of the cDNA sequence aligns only 264 basepairs from the end of the rescuing genomic fragment (accession #U57652). Based on comparison with the C. elegans translational start consensus (A/c)A(a/c)(A/C)ATG (M. Perry and W. Wood, pers. comm.), there is a possible translational start site at nucleotide 24 of the cDNA. This is the first methionine codon in-frame with the open reading frame. The next possible methione start codon is at nucleotide 418; if this were used it would imply an exceptionally long untranslated 5’ end for the message. Examination of the sequences upstream from the cDNA revealed a possible TATA box at position -185 from the putative ATG start codon (position 105 of genomic sequence). There are also three CAAT box sequences preceding the TATA box by 29-54 nucleotides (positions 47, 60 and 72). The presence and spacing of these elements correlates well with what has been observed in other eukaryotic genes (Watson et al., 1987). The TATA box differs from the consensus (TAAATA for fer-1 versus TATAAA for the consensus), but the CAAT sequences match the consensus. This would place the transcriptional start around -159 from the first ATG, with three stop codons in what would be the 5’ untranslated region.
Many C. elegans transcripts possess one of two 21 nucleotide leader sequences (called SL1 and SL2) that are trans-spliced to their 5’ ends (Blumenthal, 1995). Using a primer near the 5’ end of the fer-1 cDNA in conjunction with either an SL1-or SL2-specific primer, we obtained no amplification of the expected fragment, whereas a known trans -spliced gene was positive, so the fer-1 transcript does not appear to be trans-spliced.
fer-1 encodes a large protein predicted to be membrane-bound
Using the putative translational start, fer-1 is predicted to encode a 2,034 amino acid protein with a molecular mass of 235 kDa (Fig. 6). This protein, designated FER-1 according to standard nomenclature, is rich in charged residues (31%) and has a predicted pI of 8.5. The charged residues are distributed throughout the protein such that no particularly acidic or basic regions are observed. A hydrophilicity plot shows that FER-1 is hydrophilic except for a single 34 amino acid hydrophobic region near the C terminus which could act as a transmembrane domain (Fig. 7). The 19 amino acids from 2,001 to 2,020 are all hydrophobic and are predicted to form an alpha helix by the Robson-Garnier algorithm (Garnier et al., 1978). Other notable features of the protein are a threonine/serine/lysine rich region near the N terminus (a.a. 51-70) and a cysteine-rich region near the C terminus (a.a. 1,852-1,876), which has 10 of the 42 cysteines present in the protein, in a 25 amino acid stretch (Fig. 6). Searches of the data bases with the sequences of these regions revealed no significant similarly in other proteins. Searches of FER-1 for functional motifs from the PROSITE database using the GCG MOTIFS program and for conserved protein domains using PROFILESCAN (Devereux et al., 1984) revealed no matches with FER-1 other than highly non specific motifs.
FER-1 shows similarity to predicted human proteins
Comparison of FER-1 to other proteins in the databases using BLASTP (Altschul et al., 1990) reveals little similarity to proteins of known function. Similarities are found, however, by using TBLASTN to search the predicted proteins encoded by partial cDNA sequences (expressed sequence tags or ESTs) from C. elegans and humans. Two C. elegans ESTs, clones YK4A12 and YK5D11, likely represent fer-1 cDNAs since their sequences match that of the fer-1 cDNA almost exactly. There are 10 human ESTs whose predicted proteins are similar to FER-1 with a probability of occurring at random of less than 10−6. Alignment among these show they represent seven different genes. The three best alignments of these human proteins with FER-1 are shown in Fig. 8. Fig. 8A shows a protein sequence represented by three ESTs, two from infant brain and one from pancreatic islet cells. It is 37% identical and 82% similar over an 81 amino acids region from position 1,429-1,510 in the FER-1 sequence. Included in this region is mutation fer-1(b232ts) although the altered amino acid is not one conserved in the human protein. Fig. 8B shows a protein sequence from an adult heart cDNA clone with 33% identical and 58% similar amino acids in a region of 81 amino acids, FER-1 1,634-1,715. This region also contains a mutation, fer-1(hc80), which alters a glutamic acid that is conserved between FER-1 and the human protein. Fig. 8C shows a third protein sequence encoded by two identical cDNA clones from human breast tissue. In a region of 83 amino acids, FER-1 1,7231,804, 43% are identical and 60% similar. There are no identified fer-1 mutations in this region, but fer-1(hc24ts) is nearby at position 1,810. A third human EST (H71264) also encodes a protein that matches in this region, but it is not identical to the two breast proteins shown. Three other ESTs (HSC1rh111 from infant brain, R62744 from placenta, and W47661 from senescent fibroblasts) all encode similar but distinct proteins matching FER-1 1,935-2,010, with about 35% identity in a 6075 amino acid region.
fer-1 related genes present in the C. elegans genome
The C. elegans genome sequence project has recently sequenced fer-1. The fer-1 sequence in the database agrees with ours except for an eight nucleotide insertion in intron 10, containing a restriction site which is likely to be a cloning artifact. Another gene similar to the 5’ half of fer-1 overlaps cosmids T05E8 and F27C1 on chromosome I next to the dpy-5 gene.
Identification of fer-1 mutations
We determined the nucleotide changes associated with five of the 10 existing fer-1 mutations by sequencing fer-1 fragments amplified by PCR from homozygous fer-1 mutant worms. The mutations identified (hc13ts, hc24ts, hc80, hc136, b232ts) are all EMS-induced missense mutations and are all G/C to A/T transitions as expected for EMS mutagenesis (Anderson, 1995). These mutations all cause changes in the predicted FER-1 amino acid sequence (Figs 5, 6). We sequenced approximately 5 kb from the 3’ end offer-1 for each mutant and these mutations were the only changes detected. We did not locate the remaining five unidentified mutations (hc1ts, hc47, hc83ts, hc91ts and eb7); they probably reside in the portion of the gene not examined.
The positions of two of the identified mutations strengthen the importance of the sequence similarities observed between FER-1 and the predicted human proteins. As seen in Fig. 8B, the hc80 mutation changes a conserved glutamic acid to lysine in the 13 amino acid stretch FER-1 shares with the adult heart protein. The b232ts mutation causes an amino acid change in the region shared between FER-1 and the infant brain protein, but does not change a conserved residue (Fig. 8A). This result suggests the sequences similar between FER-1 and the predicted human proteins represent regions functionally important to these proteins.
DISCUSSION
During spermiogenesis in C. elegans, the membranous organelles (MOs) fuse with the plasma membrane, a pseudopod is extended, and directed surface membrane flow is established. Mutations in the gene fer-1 disrupt all three of these events. While fer-1 mutant spermatids appear normal and initiate spermiogenesis, the resulting spermatozoa are defective. The pseudopodia are short and stubby, although the pseudopodial projections appear normal (Ward et al., 1981), and the cells never establish the ordered surface membrane flow observed on wild-type spermatozoa (Roberts and Ward, 1982). fer-1 mutant spermatozoa are defective for MO fusion, although the MOs are oriented properly at the plasma membrane (Ward et al., 1981). The phenotype offer-1 mutant spermatozoa indicates that pseudopod initiation and elongation are distinct steps and that only elongation is dependent on MO fusion.
What role does the fer-1 gene product play during spermio-genesis? The simplest hypothesis is that fer-1 mutations block MO fusion which leads to the pseudopod and membrane flow defects. MO fusion does precede pseudopod extension, as this hypothesis demands, when spermiogenesis is initiated in vitro by treatment with the weak base triethanolamine (TEA) or the ionophore monensin (Nelson and Ward, 1980; Shakes and Ward, 1989). Since spermatozoa formed after TEA treatment are fertile as assayed by artificial insemination (LaMunyon and Ward, 1994), this order probably represents the normal order of events.
We have cloned and sequenced fer-1 in an attempt to determine the function of its gene product. DNA transformation rescue experiments and sequence analysis reveal fer-1 to be at least 8.6 kb. The cDNA and coding region of fer-1 hybridize to a single 6.2 kb sperm-specific transcript whose expression is confined to primary spermatocytes as determined by in situ hybridization. Sperm-specific expression is in agreement with analysis of fer-1 mutations, which only affect the sperm. fer-1 is predicted to encode a 235 kDa basic protein with a hydrophobic domain near the C terminus that may make it an integral membrane protein. This putative transmembrane domain is the only recognizable functional motif in FER-1. The molecular analysis reported here, in conjunction with previous examination of fer-1 mutant sperm, suggests FER-1 may play a direct role in MO-plasma membrane fusion. fer-1 mutations disrupt this fusion event without discernible effects on the structure of the MOs or their positioning at the plasma membrane (Ward et al., 1981).
Membrane fusion is an essential process for nearly all organisms and has been studied in numerous biological systems including yeast, mammalian cells, and animal viruses (Ferro-Novick and Jahn, 1994; Kaiser and Shekman, 1990; Mellman, 1995; Rothman and Orci, 1992; White, 1992; Whiteheart and Kubalek, 1995). Proteins involved in eukaryotic vesicle fusion are highly conserved from yeast to humans (Ferro-Novick and Jahn, 1994). Comparison of the predicted FER-1 protein sequence with the databases shows that it shares no primary sequence similarity to any known eukaryotic membrane fusion proteins. Based on the high degree of conservation observed for these proteins, FER-1 is clearly not a member of these protein families. If FER-1 is involved in membrane fusion, one possibility is that it is a eukaryotic analog of the viral fusion proteins. That FER-1 lacks sequence similarity to known and proposed viral fusion proteins is not surprising since the viral proteins appear to be highly divergent but share specific structural features (White, 1992). One is that they are all integral membrane proteins anchored into the viral membrane by a C-terminal transmembrane domain with >85% of the protein mass on the outside of the virus (White, 1990, 1992). Another hallmark of these proteins is the presence of a stretch of 18-36 hydrophobic amino acids called the fusion peptide. There are examples of eukaryotic proteins involved or implicated in membrane fusion that resemble the viral proteins (White, 1990). The best characterized of these is the guinea pig protein PH-30 (Blobel et al., 1992), which facilitates sperm-egg fusion. There is also a putative viral-like fusion protein in C. elegans called ADM-1 (Podbilewicz, 1996). FER-1 appears to possess a C-terminal transmembrane domain that could anchor it into one of the cellular membranes with 98% of the protein on one side of the membrane. This is reminiscent of the viral fusion proteins, but so far nothing resembling a fusion peptide has been recognized. It is also worth noting that v-SNARES, key proteins involved in cytoplasmic fusion events such as synaptic vesicle release, are also anchored by their C-terminal, with the bulk of their mass facing the cytoplasm (Südhof, 1995).
If our hypothesis that FER-1 is acting as a fusion protein is correct, it predicts that FER-1 should be localized to the interface between the MO and plasma membranes. We have attempted to raise antibodies against FER-1 using both synthetic peptides and a ∼90 kDa fragment expressed in E. coli for use in immunolocalizing the protein. These antibodies have not given reproducible sperm-specific staining so they presumably must cross react with other antigens. They do show staining of the MOs in spermatids, but we do not have deletion mutants that would allow us to determine if this staining represents FER-1. The presence of other staining patterns has made these results uninterpretable.
There are at least seven different genes expressed in the human genome that encode proteins with significant similarity to FER-1. These genes are all represented by cDNAs isolated from a wide range of tissues: brain, heart, breast, placenta, pancreas, fibroblasts and fetal liver or spleen. All of these genes were previously unidentified, so nothing is known about the function of the proteins they encode. Also, since they have only been identified by short cDNA sequences, the size of the full length proteins and the position of the domains similar to FER-1 are not known. It may be that some of the proteins will share much larger regions of similarity with FER-1 once their complete sequences are known. The regions of similarity are all in the C-terminal portion of FER-1, from amino acids 1,400 to 1,800. Two out of three of the regions of similarity between FER-1 and human proteins include FER-1 mutations, strengthening the interpretation that the regions of similarity represent functionally conserved domains. This suggests that further elucidation of FER-1 function may help determine the function of these human proteins.
ACKNOWLEDGEMENTS
We thank Paul Muhlrad, Craig LaMunyon, Jeremy Nance, Andrea Wellington and Karen Achanzar for critical reading of this manuscript; Paul Muhlrad and Marilyn Kramer for their assistance in preparation of this manuscript; and Dr Benjamin Podbilewicz for providing us with a copy of his manuscript prior to publication. This work was supported by National Institute for General Medical Sciences Grant 25243. W.E.A. also received support from Heart, Lung and Blood Postdoctoral Training Grant HL07249.