Infertility in female mice with an oocyte-specific knockout of GPI-anchored proteins

The molecular basis of mammalian sperm-egg fusion is being studied by identifying the interactions of adhesion and potential fusion proteins on mouse sperm and egg. A current model for sperm-egg fusion involves the interaction of sperm ADAM family members (fertilin and cyritestin) with an egg integrin (α₆β₁) via the ADAM disintegrin domain (Cuasnicu et al., 2001; Primakoff and Myles, 2002; Talbot et al., 2003). This ADAM-integrin model has been supported by in vitro experiments, but the in vitro evidence has been contradicted by gene-knockout studies. Sperm from male mice lacking cyritestin (Shamsadin et al.,1999) fused with eggs at 100% of the wild-type rate; sperm lacking fertilin β (Cho et al.,1998) fused at ∼50% of the wild-type rate, and double-knockout sperm, lacking both cyritestin and fertilin β, also fused at ∼50%(Nishimura et al., 2001). Furthermore, in vitro studies using culturedα ₆-integrin-knockout eggs indicated that deletion ofα ₆ had no effect on sperm-egg binding or fusion(Miller et al., 2000), and recent results indicate that none of the mouse integrins are required(He et al., 2003). Thus, ADAMs and integrins may not have the role previously attributed to them.

A number of surface proteins on sperm cells, including DE(Cohen et al., 2000) and equatorin (Toshimori et al.,1998), have been proposed to act in gamete fusion, but their role in this process has not been thoroughly studied. CD9, a member of the tetraspanin family of proteins present on eggs, has been shown to be essential for sperm-egg fusion by gene knockout(Kaji et al., 2000; Le Naour et al., 2000; Miyado et al., 2000). The exact role of CD9 on the oocyte plasma membrane is currently under investigation. The large extracellular loop of CD9 seems to have functionally significant cis interactions with other egg-surface proteins, and it seems to have an active site in its large extracellular loop that acts in gamete fusion(Zhu et al., 2002).

Another group of egg surface proteins implicated in the gamete fusion process is the lipid-linked, glycosylphosphatidylinositol-anchored proteins(GPI-APs). Coonrod and coworkers reported that treatment of mouse eggs with the enzyme PI-PLC (phosphatidyl inositol-specific phospholipase C) released two GPI-APs (∼35-45 kDa and ∼70 kDa), and caused a reduction in sperm-egg binding and a strong block of sperm-egg fusion(Coonrod et al., 1999). GPI-APs are a functionally diverse group of proteins that includes adhesion molecules,receptors, complement regulators, enzymes and signaling molecules(Ferguson and Williams, 1988; Hooper, 1997; Low, 1989). Aside from the C-terminal lipid linkage another common characteristic they share is a high degree of localization to cholesterol and sphingolipid-rich microdomains found within the plasma membrane of every cell type examined(Anderson and Jacobson, 2002; Brown and London, 2000).

The finding that PI-PLC release of oocyte surface proteins blocks gamete fusion could mean that one or both of these proteins (∼35-45 kDa and∼70 kDa) has a role in sperm-egg fusion(Coonrod et al., 1999). However, PI-PLC treatment could also have artifactual effects on membrane fusion including: (1) the PI-PLC used was not tested by enzyme assay for contaminating protease activity; (2) PI-PLC has a second substrate,phosphatidylinositol, and its loss from the plasma membrane outer leaflet might reduce membrane fusibility; (3) alternatively, PI-PLC-catalyzed production of diacylglycerol (from phosphatidylinositol and GPI-APs) in the outer leaflet of the plasma membrane could reduce fusibility; and (4) the diacylglycerol produced in the outer leaflet might flip into the inner leaflet and initiate signaling that would block gamete fusion. In addition, in vitro experiments may show effects that can not be confirmed by in vivo tests using gene deletion (Hynes, 1996). This has clearly been the case in fertilization research, where some of the predicted roles of acrosin (Baba et al.,1994), Trp2 calcium channels(Leypold et al., 2002),galactosyltransferase (Lu and Shur,1997) and the ADAMs and integrinα ₆β₁ mentioned above have failed to be confirmed by in vivo knockout studies.

To determine whether GPI-APs have a role in fertilization in vivo we created conditional, oocyte-specific GPI-AP-knockout mice. Pig-a,phosphatidylinositol glycan class-A, encodes a subunit of an N-acetyl glucosaminyl transferase that is involved in the first steps of GPI anchor biosynthesis (Tiede et al.,2000). Complete knockout of Pig-a is embryonic lethal(Kawagoe et al., 1996);therefore the Cre/loxP recombination system was used to disrupt the gene Pig-a so that GPI-AP delivery to the plasma membrane was precluded in oocytes. Disruption of Pig-a exon 6, through flanking it with direct loxP repeats(Tarutani et al., 1997),occurred only in oocytes and not in other cells because Cre-recombinase was expressed under control of the oocyte-specific promoter ZP3 (Shafi et al.,2000). Transgenic mice carrying the ZP3-Cre gene and two copies of the loxP-flanked Pig-a gene were considered to be conditional knockouts. We have shown that egg GPI-APs are required for sperm-egg fusion by use of the conditional knockout females.

Mice carrying `floxed' Pig-a alleles(Pig-a^flox) (Tarutani et al., 1997) and a transgene encoding a GPI-anchored enhanced green fluorescent protein (eGFP-GPI)(Kondoh et al., 1999) have been described previously. The Pig-a gene is located on the X chromosome. The presence of a Pig-a^flox allele (exon 6 flanked by direct loxP sites) was detected by PCR on genomic tail DNA as described previously (Tarutani et al.,1997). DNA templates were denatured for 3 minutes at 94°C,then amplified for 30 cycles consisting of 94°C denaturation for 1 minute,annealing at 65°C for 0.5 minutes and elongation at 68°C for 0.5 minutes; this was followed by a final extension at 68°C for 8 minutes. The PCR reaction was run on a 1.2 % agarose 1× TAE (40 mM Tris-Acetate, pH 8.3, 2 mM EDTA) gel and visualized by ethidium bromide staining and UV illumination. The wild-type Pig-a allele gives rise to a 260 bp band,and the floxed allele gives rise to a 420 bp band. Homozygous females were designated Pig-a f/f, and males were designated Pig-a f/y. The eGFP-GPI transgene was detected in newborn pups by shining UV light on their skin; homozygous pups fluoresced more brightly than hemizygous pups. The ZP3-Cre transgene was detected in genomic tail DNA from ZP3-Cre mice (provided by Jamey Marth, UCSD, CA) by PCR using the same parameters as above. The 5′ primer, 5′-GGA CAT GTT CAG GGA TCG CCA GGC G-3′, located 123 bp downstream of the Cre ATG start site, and a 3′ primer, 5′-GCA TAA CCA GTG AAA CAG CAT TGC TG-3′,generated a 268 bp product.

Pig-a^flox females carrying the eGFP-GPItransgene (Pig-a f/f:eGFP-GPI) were mated with ZP3-Cre males to generate ZP3-Cre:Pig-a f/y:eGFP-GPI pups. These male mice were mated with Pig-a f/f:eGFP-GPI females to generate a conditional knockout female with the genotype ZP3-Cre:Pig-a f/f:eGFP-GPI.

Conditional knockout or wild-type C57BL/6 females, 2 to 6 months old, were individually housed for 21 days with a wild-type C57BL/6 male (Charles River)aged eight to 10 weeks. Females were then separated from the males and allowed to rest for 21 days, the average gestation period in mouse. A female was considered fertile if she gave birth to pups. Litter size was determined by counting pups.

Wild-type C57BL/6 and conditional knockout females were superovulated with 10 IU pregnant mares' serum gonadotropin (PMSG, Sigma) followed 46 to 50 hours later with 10 IU human chorionic gonadotropin (hCG, Sigma). Immediately following hCG injection a wild-type C57BL/6 male was introduced into the cage with a single female. 40 hours later the females were euthanased and the oviducts were removed. The oviducts were placed in supplemented M199, which is M199 medium (GIBCO BRL) supplemented with 3.5 mM sodium pyruvate, 100 IU/ml Penicillin, 100 ug/ml Streptomycin and containing 0.4% BSA (Fraction V, fatty acid free, Sigma). To release unfertilized oocytes and/or two cell embryos the oviducts were minced with dissecting scissors. Two-cell embryos and unfertilized oocytes were transferred through three 500 μl drops of medium then viewed by light microscopy at 20× magnification to score the number of two-cell embryos and unfertilized oocytes.

Wild-type C57BL/6 and conditional knockout females were superovulated as described above. 13 to 14 hours post-hCG injection cumulus masses were retrieved from the oviducts of euthanized females and placed in supplemented M199 medium + 0.4% BSA. The supplemented medium was pre-equilibrated overnight under light mineral oil (Fisher) in a humidified incubator with 5%CO₂ at 37°C. Cumulus cells were removed from oocytes by treatment in 300 μg/ml hyaluronidase Type I-S (Sigma) for 5 minutes followed by washing through three 500 μl drops of medium. To remove the zona, oocytes carrying a first polar body were transferred to a 100 μl medium drop containing 30 μg/ml α-chymotrypsin (Sigma) and incubated for 3 minutes. Oocytes were transferred to a fresh medium drop and then passed several times through a narrow bore (∼80 μm), hand-pulled Pasteur pipette to release eggs from the partially digested zona. Zona-free oocytes were washed through two 100 μl drops of medium and placed in a fresh drop to recover for 3 hours. Recovered oocytes were loaded with 4′,6′Diamidino-2-phenylindole dihydrochloride (DAPI) to label DNA by incubating the cells for 12 minutes at 100 μg/ml, 37°C, 5% CO₂ followed by washing through three drops.

Sperm were isolated from 10- to 12-week-old ICR males (Charles River). The cauda epididymis and vas deferens were removed from euthanased males and placed in a 500 μl drop of supplemented M199 + 3% BSA under light mineral oil. Sperm were squeezed from the vas deferens, and the epididymis was cut in several places and placed for 15 minutes in a humidified CO₂chamber to allow sperm to swim out. Tissue fragments were removed, and the sperm diluted 10-fold into a fresh 500 μl drop to a concentration of 1-5×10⁶ sperm/ml and capacitated for 3 hours at 37°C, 5%CO₂.

Capacitated sperm were diluted 10-fold to a final concentration of 1-5×10⁵ sperm/ml in a 100 μl drop of supplemented M199 +0.4%BSA along with DAPI-loaded, zona-free oocytes. Gametes were co-incubated for 40 minutes at 37°C, 5% CO₂. Sperm-egg complexes were washed through one 100 μl medium drop and loaded onto a microscope slide to score for the fertilization rate (FR; percentage of eggs fused with at least one sperm), fertilization index (FI; total number of sperm fused per total number of eggs) and sperm bound per egg. The transfer of DAPI from pre-loaded eggs to sperm was used to score sperm-egg fusion.

Zona-free oocytes from wild-type C57BL/6, from Pig-a f/f:eGFP-GPI, and from conditional knockout ZP3-Cre:Pig-a f/f:eGFP-GPI females were collected as described above. To assess surface expression of GPI-APs, oocytes were probed with a rabbit polyclonal anti-GFP antibody (Abcam Limited, ab290) to detect the presence of eGFP-GPI. Eggs were washed through four 100 μl drops of PBS(137 mM NaCl, 2.7 mM KCl, 4.3 mM Na₂HPO₄, 1.4 mM KH₂PO₄, pH 7.4) + 3% BSA (PBS/BSA), transferred to 100μg/ml rabbit anti-GFP in PBS/BSA and incubated at room temperature for 1 hour. Eggs were washed through three 100 μl drops of PBS/BSA, transferred to 20 μg/ml Alexa Fluor® 568 goat anti-rabbit secondary antibody(Molecular Probes) and incubated for 30 minutes at room temperature. Eggs were washed through three drops of PBS/BSA and placed under a raised coverslip on a microscope slide and viewed by laser scanning confocal microscopy (model LSM 410, Carl Zeiss). The 568 laser line was used to excite the Alexa Fluor®-568-conjugated secondary antibody.

Essentially the same protocol was used to test a set of antibodies to known mammalian GPI-APs (anti-CD24, -CD48 and -Qa-2 from Pharmingen, San Diego, CA;anti-CD55, GPC3, uPAR, Gas1 antibodies from Santa Cruz Biotechnology, Santa Cruz, CA) on zona-free oocytes. In the case of CD55 the first antibody was 100μg/ml polyclonal goat anti-mouse CD55 (Santa Cruz Biotechnology) and the secondary antibody was 20 μg/ml Alexa Fluor® 488-donkey anti-goat(Molecular Probes). PBS + 1% polyvinyl alcohol (PBS/PVA) was used as the buffer for antibody incubation and all washes. As a control for primary antibody binding, wild-type oocytes were incubated with an irrelevant Goat IgG(Zymed) at 100 μg/ml. To determine if oocyte CD55 was PI-PLC sensitive,wild-type oocytes were treated with 1 U/ml PI-PLC (kind gift of Martin Low,Columbia University) in supplemented M199 + 0.4% BSA for 30 minutes, 37°C,then washed through three 100 μl droplets of medium prior to antibody labeling.

Zona-intact oocytes were retrieved as described above and incubated with or without 1 U/ml PI-PLC in PBS. The oocytes were incubated for 30 minutes at room temperature, after which cells were removed, washed through four 100μl drops of PBS/PVA, then transferred to an Eppendorf tube with SDS sample buffer (2% SDS, 125 mM Tris-HCl, pH 6.8, 20% glycerol and 0.2% bromophenol blue). The samples were heated, separated on 10% SDS-PAGE and transferred to PVDF for western blotting. The PVDF membrane was blocked for 1 hour at room temperature in TBST (50 mM Tris-HCl pH 7.3, 500 mM NaCl, + 0.5% Tween 20) + 5%non-fat milk. The membrane was probed for 1 hour at room temperature in the same blocking solution + 40 ng/ml goat anti-mouse CD55. HRP-conjugated donkey anti-goat (Santa Cruz Biotechnology) at 2 ng/ml was incubated with the membrane in blocking solution for 1 hour. The membrane was washed three times for 10 minutes in TBST at room temperature then developed in Pierce's WestFEMTO chemiluminescent reagent following the manufacturer's instructions.

To study the in vivo role GPI-APs play in sperm-egg plasma membrane binding and fusion, we generated conditional Pig-a-knockout mice in which only oocytes are affected. Disruption of the gene Pig-a, which encodes an N-acetyl glucosamine transferase required in the first steps of GPI biosynthesis, results in the absence of GPI-APs from the cell surface(Kawagoe et al., 1996) and embryonic lethality at 9 days post coitus in mouse(Nozaki et al., 1999). To direct oocyte-specific disruption of the Pig-a gene, the Cre/loxP recombination system was used to delete Pig-a exon 6, which was flanked by direct repeats of the loxP sequence. Cre-recombinase,which catalyses the excision, was put under control of the oocyte-specific ZP3 promoter. ZP3-Cre-mediated excision of Pig-a exon 6 is expected to occur only in oocytes and to inactivate the gene product(Tarutani et al., 1997).

Appropriate matings allowed us to obtain conditional knockout female mice designated ZP3-Cre:Pig-a f/f:eGFP-GPI (Fig. 1). These females carry the ZP3-Cre transgene, two Pig-a^flox alleles (Pig-a f/f) and an internal control transgene encoding a GPI-anchored eGFP(Kondoh et al., 1999). To test fertility, 10 conditional knockout females were mated for 21 days with a wild-type C57BL/6 male (one female housed with one male) and then rested another 21 days during which time no visible pregnancies or births were observed (Table 1). All 10 wild-type control females became pregnant and gave birth to an average of 6.2±2.5 pups within 23 days of males being introduced.

One possible explanation for the infertility found in conditional knockout mice is that the absence of GPI-APs from the surface of developing oocytes prevents maturation of oocytes. To determine whether mature oocytes could be retrieved, wild-type and conditional knockout mice were superovulated with serum gonadotropins, and zona-intact oocytes were retrieved from the ampullae of treated females as described in Materials and Methods. The number of oocytes and percentage of mature oocytes having a first polar body from conditional knockouts and wild-type animals were essentially the same(Table 2).

To determine the extent to which GPI-AP expression was affected, oocytes were retrieved from knockout females carrying the eGFP-GPI transgene and analyzed by confocal microscopy to detect expression of eGFP-GPI on the egg plasma membrane. eGFP-GPI, detected with a GFP-specific antibody, was present on the surface of wild-type oocytes but was absent from the surface of conditional knockout oocytes (Fig. 2), which indicates successful Pig-a disruption.

Although they produce normal numbers of superovulated, mature oocytes,conditional GPI-AP-knockout females never became visibly pregnant. To test if fertilization was occurring in vivo we mated gonadotropin-treated females and scored the number of two-cell embryos as a measure of fertilization. A male was introduced into the cage with a single superovulated female immediately post-hCG injection as described in Materials and Methods and left for 40 additional hours. Females were subsequently euthanized, then cells(unfertilized eggs or two-cell embryos) were retrieved from the oviducts of mated wild-type and conditional knockout females. From three wild-type females 83% of the 52 healthy cells retrieved were two-cell embryos. However, from a total of five conditional knockouts 74 healthy, unfragmented cells were retrieved, and among these an average of only 1.3% were two-cell embryos(Fig. 3). In addition, multiple sperm were detected in the perivitelline space of several Pig-a^–/– eggs. This indicates that there was no block to multiple sperm traversing the zona and entering the perivitelline space, a block that does occur if a sperm fuses with the egg.

Since unfertilized eggs with perivitelline sperm were observed and two-cell embryos were almost absent, it appears that sperm-egg fusion is defective in conditional knockout females when females are mated. To test for a fusion defect in GPI-AP-knockout eggs, zona-free oocytes obtained from conditional knockout and wild-type females were inseminated with wild-type sperm for 40 minutes. The fertilization rate was about eight-fold lower and the fertilization index about nine-fold lower in GPI-AP^–/–eggs compared to the wildtype (Fig. 4). Sperm binding to the knockout eggs was not significantly different from the control (3.9±1.8 and 2.4±1.9 wil type versus knockout; P=0.09). These findings indicate that when GPI-APs are not expressed on the surface of mouse eggs, the eggs are defective in fusing with sperm.

PI-PLC treatment of surface-biotinylated mouse eggs releases an ∼35-45 kDa and a 70 kDa protein detectable with avidin(Coonrod et al., 1999). We sought to identify these bands or other egg-surface GPI-APs initially by using commercially available antibodies to known mammalian GPI-APs. One of the commercial antibodies identified an egg-surface GPI-AP as CD55 in indirect immunofluorescence and western blot experiments. CD55 is a complement regulatory protein also known as decay accelerating factor, DAF. Zona-free wild-type mouse oocytes were stained with polyclonal goat anti-CD55(Fig. 5A). Pretreatment of eggs with PI-PLC, which releases GPI-APs from the cell surface, diminished staining(Fig. 5B), indicating that oocyte CD55 is PI-PLC sensitive. By western blotting, a single ∼70 kDa protein band was recognized in mouse eggs(Fig. 6). PI-PLC sensitivity was also demonstrated in this blot by the decrease in staining intensity of the ∼70 kDa band when eggs were pre-treated with PI-PLC. Unfortunately,when supernatants from PI-PLC-treated eggs were analyzed, the anti-CD55 antibody was unable to detect protein, suggesting that the epitopes are less available or altered after CD55 release.

We created oocyte-specific Pig-a-knockout female mice to test the requirement of GPI-APs in fertilization. In fertility trials, we found that conditional knockout females are infertile and fail to produce fertilized eggs when mated with wild-type males. Infertility is not caused by defects in egg development, ovulation or the ability to mate with a partner. Mature oocytes with normal morphology, but lacking surface expression of GPI-APs, are retrieved from conditional knockout females in equivalent numbers to wildtype. Sexual behavior is normal in knockout females: when they are mated with a male, multiple sperm reach the site of fertilization, the oviduct, and cross into the perivitelline space of knockout eggs. Taken together, the gene deletion evidence shows that egg GPI-APs are required for fertilization.

The in vivo data suggest that the female infertility observed is the result of a block to fertilization at the level of sperm-egg fusion. Multiple sperm enter the perivitelline space of Pig-a^–/– eggs without causing a fusion-induced block to polyspermy. The in vitro fertilization data confirm that the defect is at the level of gamete fusion because wild-type sperm fuse poorly with Pig-a^–/– eggs.

A number of interpretations can be considered to explain why eggs lacking GPI-APs are defective in fusion. One possibility is that eggs may require their full complement of GPI-APs to maintain the correct membrane organization to be fusion-competent. Our data along with the known lack of an effect of the CD55 knockout on fertility indicate that this possibility is not the case. The GPI-AP CD55 was identified on mouse eggs. However, in previous studies CD55-knockout mice have been generated, and they do not show altered fertility(Sun et al., 1999). Therefore,even though the CD55-knockout female mice lack the full complement of GPI-anchored proteins, they have normal fertility.

A second possibility is that at least some GPI-anchored proteins must be present to support the interaction of proteins in organized plasma membrane lipid domains (e.g. rafts). In some cell types when all GPI-APs are absent the cells still contain rafts, indicating that GPI-APs are not indispensable structural elements of these microdomains(Abrami et al., 2001). However,the absence of all the egg GPI-APs may alter the normal lipid domain composition and may change the repertoire of protein-protein interactions in lipid domains. The lipid domain is an important functional unit for cell signaling molecules, for cell activation, adhesion and viral infectivity. For example, in T cell receptor (TCR)-mediated signaling and activation, the function of the TCR and the complex of proteins associated with it is dependent on lipid domain integrity and the presence of GPI-APs(Montixi et al., 1998; Romagnoli and Bron, 1997). Disruption of lipid domains by cholesterol depletion or lack of GPI-APs in T cells results in decreased signaling. T cells lacking GPI-APs have decreased phosphorylation downstream of TCR activation probably because src family kinases found in lipid raft fractions have decreased activity in the absence of GPI-APs (Romagnoli and Bron,1997). Lipid rafts may also serve as platforms for viral entry into some cell types (Dimitrov,2000), and disruption of these domains prevents entry of a variety of pathogens (Campbell et al.,2001). Thus, the absence of egg GPI-APs from lipid domains might block egg signaling or egg functions that normally result in membrane fusion through steps about which nothing is currently known.

A final explanation is that a single egg GPI-AP has a specific, required function in sperm-egg fusion. Biotinylation of the egg surface followed by PI-PLC treatment and detection with avidin reagents revealed GPI-APs of∼70 kDa and ∼35-45 kDa (Coonrod et al., 1999). We found that CD55 is a GPI-AP on the egg surface,which has a M_r ∼70 kDa and may be the biotinylated 70 kDa protein. However, CD55 apparently does not have a required function in fertilization as CD55-knockout mice have normal fertility(Sun et al., 1999). To evaluate a possible specific function in gamete fusion for a single egg GPI-AP, it will be necessary to characterize and identify the remaining∼35-45 kDa GPI-AP or other so far undetected GPI-APs that may be of lower abundance and/or poorly biotinylated.

Since egg GPI-APs and CD9 are now the two known gene products whose ablation blocks sperm-egg fusion, it is worth considering if and how they may interact. Tetraspanins in general and CD9 in particular are not concentrated in the kinds of lipid domains that are enriched for GPI-APs(Claas et al., 2001). Although CD9 is known to form associations with many other plasma membrane proteins(e.g. Ig superfamily, membrane-anchored growth factors and integrins), none of the known CD9 partners has a GPI anchor. These preliminary ideas suggest the GPI-APs and CD9 act in distinct segments of a fusion pathway in which additional components remain to be revealed.

We would like to thank Jamey Marth for providing ZP3-Cre mice, Masaru Okabe for making Pig-a^flox mice available and Martin Low for kindly giving PI-PLC. We would also like to thank Jowell C. Go for expending valuable time genotyping mutant mice. This work was supported by the National Institutes of Health grant HD16580 and the National Institutes of Health Fertilization and Early Development training grant.