Unveiling a novel function of CD9 in surface compartmentalization of oocytes

In multiple phases of mammalian gamete fusion, IZUMO1 (Inoue et al., 2005) and its receptor JUNO (also known as IZUMO1R) (Bianchi et al., 2014) are the well-documented first pair of gamete recognition factors of the spermatozoon and the ovum, respectively. We, as well as another group, have recently clarified the tertiary structures of the human IZUMO1-JUNO complex at atomic resolution (Aydin et al., 2016; Ohto et al., 2016). We have also shown that the initial gamete recognition step can be reconstituted in cultured cells expressing the Izumo1 gene with oocytes (Inoue et al., 2013). As the cultured cells solely expressing the Izumo1 gene lacked membrane fusion activity, IZUMO1-JUNO interaction is considered to be a recognition step before membrane fusion (Inoue et al., 2015); however, the detailed molecular mechanisms underlining gamete fusion remain elusive.

The oocyte surface (oolemma) in mammals is mainly divided into two areas: the microvilli-devoid region located on the surface of the cortical actin cap region, which anchors meiosis metaphase II chromosomes (Uraji et al., 2018), and the microvilli-rich region, in which the spermatozoon binds to the oocyte (Yanagimachi, 1994). Cluster of differentiation 9 (CD9), a member of the tetraspanin family of proteins, is required for accomplishing gamete fusion, as female CD9 mutants have been found to exhibit severely reduced litter size and gamete fusion ability (Kaji et al., 2000; Le Naour et al., 2000; Miyado et al., 2000). In 2007, Runge et al. first reported that CD9-disrupted oocytes have microvilli of altered size, thickness and density (Runge et al., 2007), and another group confirmed that CD9 deletion increased microvilli thickness (Benammar et al., 2017). Although there are discrepancies in the reports of the effects of CD9 deletion on microvilli density (Benammar et al., 2017; Miyado et al., 2008; Runge et al., 2007), oocyte CD9 is at least regarded as an indispensable factor for the formation of normal microvilli.

CD9 is known to interact with various proteins, such as integrins, and forms a dynamic complex termed the ‘tetraspanin web’ (Boucheix and Rubinstein, 2001). In particular, oocyte CD9 is strongly associated with CD315 (EWI-F; PTGFRN) and IGSF8 (EWI-2) (Runge et al., 2007); however, IGSF8 is not functional in the fertilization (Inoue et al., 2012). IGSF8 and CD315 bind directly to ezrin, radixin and moesin (ERM) through their cytoplasmic tails, finally linking to the actin filaments of the microvillous core (Sala-Valdés et al., 2006). Intriguingly, the CD9 homologues CD81 and CD82 regulate microvillus morphogenesis positively and negatively, respectively, by altering cortical actin cytoskeletal organization (Bari et al., 2011).

Another important aspect of CD9 is its association with fertilization. Utilizing adhesion force measurements, Jegou et al. reported that strong adhesion activity of the oocyte to the spermatozoon through the tetraspanin web to trigger sperm-egg fusion is impaired in CD9-null oocytes, whereas total frequency of gamete adhesion is increased (Jegou et al., 2011). In addition, Miyado et al. demonstrated that exosome-like CD9-containing vesicle release from an oocyte leads to spermatozoon fusion-facilitating activity (Miyado et al., 2008), although this observation may be controversial because their experiments have been found to be unreproducible by two independent research groups (Barraud-Lange et al., 2012; Gupta et al., 2009).

In the present study, we provide evidence indicating that a newly discovered CD9 physiological function enables exofacial exclusion of glycosylphosphatidylinositol (GPI)-anchored proteins (GPI-APs) from the cortical actin cap region in mouse oocytes.

We first examined the localization of CD9 and JUNO in mouse eggs using α-CD9 (MZ3) or α-JUNO (TH6) monoclonal antibodies 15 min after insemination (Fig. 1). As IZUMO1 relocates from the acrosomal cap to the whole head of the spermatozoa after acrosome reaction, acrosome-reacted (AR) spermatozoa were detected by α-IZUMO1 staining of the whole sperm head (Inoue et al., 2005; Inoue and Wada, 2018; Satouh et al., 2012). Both JUNO and CD9 on the egg were found to be concentrated in regions at which spermatozoa labelled with α-IZUMO1 antibody were bound (Fig. 1, arrows), which is consistent with previous reports (Inoue et al., 2015; Kaji et al., 2002; Ravaux et al., 2018). The crescent-shape staining pattern of IZUMO1 in Egg #1 shows a state occurring within 5 min of gamete fusion (Fig. 1, arrowheads) (Satouh et al., 2012). Interestingly, in the area where the sperm-egg fusion occurred, CD9 was still fully present, whereas JUNO had partially disappeared from the surface of the oocyte in concentric circles (Fig. 1, arrowheads). Regarding JUNO, this is a response observed immediately after fusion with a spermatozoon, implying that the polyspermy block system had been launched (Bianchi et al., 2014). In Egg #2, fertilization appeared to be at a later stage because the nucleus of the fused spermatozoon started swelling and IZUMO1 was almost absent (Fig. 1, middle panels, arrowheads). Profiles of the fluorescence intensity of CD9 and JUNO indicated that, although CD9 was still present in both eggs, Egg #2 entirely lost JUNO from around the gamete fusion site, whereas JUNO was still present in the Egg #1 surface, albeit at very low levels (Fig. 1, long arrows, left bottom panels). These results suggest that the molecular dynamics of CD9 and JUNO are differentially regulated, immediately following sperm-egg fusion.

In order to analyze how CD9 and JUNO cooperate in fertilization, we generated CD9 and JUNO knockout mice using the CRISPR/Cas9 system. We designed a single guide RNA (sgRNA) to target the upstream domain of the second large extracellular loop for CD9, known as the functional domain (Zhu et al., 2002), or the initial methionine codon for the Juno gene (Fig. 2A). We then injected pX330-sgRNA plasmids into the pronuclei of fertilized eggs (Mashiko et al., 2013). Consequently, we obtained mouse lines in which 10 and 31 bases were deleted in CD9 and JUNO genomic DNA, respectively (Fig. 2A, red rectangles). Loss of CD9 or JUNO proteins in each knockout mouse line was confirmed by western blot analysis with α-CD9 (EM-04) or α-JUNO (12A5) monoclonal antibodies, respectively (Fig. 2B, Fig. S4). When expression levels of CD9 or JUNO proteins were examined from 20 eggs obtained from three independent mice, there was no significant difference in protein expression. JUNO showed smeared bands in the SDS-PAGE, implying that JUNO oocytes acquired post-translational modification (Fig. 2B).

We next performed an in vitro fertilization (IVF) assay to confirm the failure of gamete fusion, as previously reported (Bianchi et al., 2014; Kaji et al., 2000; Le Naour et al., 2000; Miyado et al., 2000). Both Cd9^−/− and Juno^−/− oocytes showed almost no fertilized eggs, whereas the fertilization rate of the wild-type oocytes was nearly 100% (Fig. 2C).

Next, we performed immunostaining analysis in living Cd9^−/− and Juno^−/− oocytes with specific monoclonal antibodies (Fig. 2D). CD9 or JUNO proteins were not observed on the surface of the Cd9^−/− or Juno^−/− oocytes, respectively (Fig. 2D, middle and bottom panels), although both proteins were detected on the surface of the wild-type oocytes (Fig. 2D, upper panels). Importantly, we noticed that in the Cd9^−/− oocytes, JUNO localization was abnormal, i.e. although JUNO was seen in the cortical actin cap region lacking microvilli, as well as the microvilli-rich region, it was nearly absent from the cortical actin cap region in the wild-type oocytes (Fig. 2D, arrows).

We previously showed that sperm-egg recognition/adhesion can be reconstituted in cultured cells expressing the Izumo1 gene with oocytes (Inoue et al., 2013). To evaluate the IZUMO1 binding ability of oocytes lacking CD9 or JUNO, we first measured the interaction of each oocyte to IZUMO1-PA- or IZUMO1-mCherry-expressing COS-7 (African green monkey kidney fibroblast-like cell line) cells. As shown in Fig. 3A, the Cd9^−/− oocytes showed clear binding to IZUMO1-PA expressed in COS-7 cells, whereas bound cells were almost absent in Juno^−/− oocytes, although the number of associated cells was reduced compared with those in the wild-type oocytes. The characteristic clustering of dimeric IZUMO1-mCherry at the contact site was observed as reported in a previous study (Fig. 3B, asterisks; Inoue et al., 2015). In Cd9^−/− oocytes, however, the strong adhesion activity between the spermatozoon and oocyte is impaired (Jegou et al., 2011), although IZUMO1 recruitment kinetics is accelerated at the contact site (Chalbi et al., 2014). Thus, the lack of CD9 likely affects the tight binding phase of the IZUMO1 molecule resulting in the reduction of the number of binding cells. Next, to further characterize the IZUMO1 and JUNO interaction, recombinant IZUMO1-FC fusion protein (IZUMO1-FC) was produced in 293T cells. Fig. 3C shows that 5 μg ml⁻¹ IZUMO1-FC bound to both wild-type and Cd9^−/− oocytes; however, no binding was observed with Juno^−/− oocytes because of the lack of IZUMO1-JUNO interaction. During this experiment, recombinant IZUMO1 staining was observed as densely clustered patches on the surface of the oocytes (Fig. 3C), which appeared to represent the bundled microvillous regions or more complex structures. Interestingly, IZUMO1-FC failed to bind to the cortical actin cap region, despite the obvious JUNO localization in the Cd9^−/− oocytes (Fig. 3C). Taken together, the IZUMO1-JUNO interaction is not hindered by CD9 disruption; therefore, IZUMO1-JUNO and CD9 may be independent pathways for triggering sperm-egg fusion.

We next considered the possibility that the aberrant JUNO distribution in the Cd9^−/− oocytes may be caused by an impaired cortical actin cap structure. We observed the oocyte surface ultrastructure using scanning electron microscopy (SEM). As shown in Fig. 4A, there was no obvious distortion in the images of surface structure surrounding the cortical actin cap region in the Cd9^−/− oocytes. All microvilli of Cd9^−/− oocytes were shorter and less densely packed than those of the wild-type oocytes, as reported previously (Benammar et al., 2017; Runge et al., 2007). The SEM image of Juno^−/− oocytes was not distinguishable from that of the wild-type oocytes. Consistently, phalloidin staining showed that F-actin was fully formed in this region, and there were no signs of abnormality in the Cd9^−/− oocytes (Fig. 4B). Indeed, the cortical actin cap was not disturbed by latrunculin B treatment, which inhibits the polymerization of monomeric actin, although patched distribution of CD9, integrin α6 and JUNO proteins was observed (Fig. S1). Thus, the cause of the aberrant JUNO distribution in the Cd9^−/− oocytes is unlikely to be the cortical actin cap structure itself.

Previous studies have shown that integrin α6 is most likely to be the cis-partner of CD9 through the tetraspanin web in oocytes (Boucheix and Rubinstein, 2001; Miyado et al., 2000). We therefore surmised that loss of CD9 might cause disturbance in integrin α6 localization. However, immunofluorescence experiments of the Cd9^−/− oocytes indicated that integrin α6, as well as integrin β1, CD315 and CD81, were excluded from the cortical actin cap region (Fig. 5A and Fig. S2A-D). On the other hand, CD55, a GPI-AP (Alfieri et al., 2003), was incompletely excluded from the cortical actin cap region in Cd9^−/− oocytes, with the pattern being very much similar to that of JUNO (Fig. S2E). When the fraction of oocytes with proper segregation of JUNO and integrin α6 was quantified, we found that all Cd9^−/− oocytes showed incomplete surface compartmentalization of JUNO, which invaded the cortical actin cap region from the rest of the area. This was in contrast to the integrin α6 compartmentalization that was almost the same as that in the wild-type oocytes (Fig. 5A, graph). These observations suggest that GPI-APs, which tend to be incorporated in lipid rafts on the plasma membrane, can invade the cortical actin cap region.

It is known that tetraspanins (CD9, etc.) are localized to the cholesterol-depletion-resistant membrane microdomain (non-lipid raft), whereas GPI-APs (JUNO, etc.) are present in the cholesterol-rich microdomain (lipid raft) (Buschiazzo et al., 2013). These two types of proteins appear to be difficult to come across in association with each other owing to their properties. However, this mode of interaction between two proteins in non-lipid raft and lipid raft domains has been reported previously, and it was abrogated by treatment with methyl-β-cyclodextrin (MβCD), a cholesterol-depleting reagent (Wang et al., 2007). As the treatment lowers lateral diffusion of CD9, as revealed by single molecule analysis, the frequency of molecular interaction could be reduced (Espenel et al., 2008). Thus, to determine whether the membrane raft integrity is required for regulation of CD9-mediated GPI-APs dynamics, we analyzed molecular localization under the cholesterol depletion. First, we determined optimized concentration of MβCD for live imaging of oocytes, and chose 2 mM MβCD (Fig. S3A). As seen in Fig. S3B, MβCD treatment showed normal molecular compartmentalization of JUNO in the wild-type oocytes, and did not improve exclusion of JUNO in the cortical actin cap region of the Cd9^−/− oocytes, although clustered punctate structures of CD9, integrin α6 and JUNO proteins appeared as observed for latrunculin B treatment (Fig. S1).

To further characterize this property, we performed fluorescence recovery after photobleaching (FRAP) experiments using α-JUNO antibodies labelled with a fluorophore to monitor diffusion of endogenous JUNO proteins. As expected, the recovery was almost negligible in the wild-type oocytes after photobleaching the cortical actin cap region. However, in the corresponding region of the Cd9^−/− oocytes, the recovery was significantly observed, albeit at a slower rate compared with experiments using the surface of cultured cells (Fig. 5B and Movie 1). We obtained similar results with CD55 (Movie 2). Thus, this suggests that CD9 is required to establish the proper cortical actin cap with strict filtration function that excludes GPI-APs.

Finally, we analyzed sperm-oocyte binding in the cortical actin cap region of the Cd9^−/− oocytes. In this assay, we employed IZUMO1-mCherry and acrosomal enhanced green fluorescent protein (Acr-EGFP) double transgenic spermatozoa to distinguish between acrosome intact (AI) (Fig. 5C, double positive, yellow) and AR (Fig. 5C, IZUMO1-mCherry single positive, red) spermatozoa. The distribution of bound spermatozoa was evaluated by integrin α6 localization because, even in Cd9^−/− oocytes, integrin α6 shows normal compartmentalization (Fig. 5A and Fig. S2A). Astonishingly, as seen in the images in Fig. 5C, 10 min after insemination with zona-free oocytes, the majority of the bound spermatozoa did not undergo acrosome reaction, suggesting that the plasma membrane binding factors of spermatozoa bind to oolemma before sperm-egg fusion, despite a non-physiological event. The number of spermatozoa bound to Cd9^−/− oolemma significantly decreased not only in the AI but also, more importantly in the AR spermatozoa (Fig. 5C, arrows). There was no statistical difference between number of spermatozoa bound to the cortical actin cap region in the wild-type and Cd9^−/− oocytes; however, multiple bound AI spermatozoa were rarely seen in Cd9^−/− oocytes (Fig. 5C, image of Cd9^−/− oocyte, arrowheads). AR spermatozoa bound to the cortical actin cap were exceptionally rare [wild-type: 0.23% (7/2997 bound spermatozoa), Cd9^−/−: 0.22% (3/1366 bound spermatozoa)], suggesting that JUNO, which is localized in the cortical actin cap region in Cd9^−/− oocytes, is somehow maintained in an inactive state for binding of IZUMO1 (Fig. 3C). Thus, we presume that the cortical actin cap region is functionally organized by multiple systems other than CD9 regulations, to which CD9 compartmentalization-unaffected transmembrane type plasma membrane binding factors (e.g. integrin) may be committed before IZUMO1-JUNO interaction, to exclude spermatozoa from there.

The results of the current study show that the characteristic cortical actin cap surface of oocytes is insufficiently formed in Cd9^−/− oocytes. We therefore propose that CD9 is required for regulating the molecular compartmentalization of GPI-APs on the cortical actin cap surface. Given that spermatozoa are unable to bind to the cortical actin cap surface which is adjacent to chromosomes, this unique zone on the cortical actin cap has to be completely organized through CD9 involvement before gamete recognition (Fig. 6).

The cortical actin cap region of oocytes is a less characterized structure. In the current study, although the surface images of the cortical actin cap of CD9-deficient oocytes were indistinguishable from wild-type oocytes when observed by SEM, further FRAP experiments clearly indicated that CD9-deficient oocytes lacked the property to exclude GPI-APs, JUNO and CD55, from the cortical actin cap region (Fig. 5B, Movies 1 and 2). Because, unlike the GPI-APs that are only tethered to the outer plasmalemmal leaflet, the membrane spanning proteins, such as integrin α6 and β1, CD315 and CD81, were still excluded from the cortical actin cap region of Cd9^−/− oocytes (Fig. S2), the unknown outer leaflet domains other than the lipid raft have to participate actively in the surface compartmentalization through the commitment of CD9.

Nucleation of actin in the cortex is known to be generated by Arp2/3 (Actr2/3) formed by the Ran^GTP gradient via the Cdc42-N-WASP or Rac-WAVE pathway (Yi et al., 2013). There are no reports that CD9 is involved in either pathway. In turn, incomplete exclusion of GPI-APs indicates that the exofacial surface in the cortical actin cap region is not as tightly packed as that in the wild-type; thus, diffusion of molecules on the outer plasmalemmal leaflet is less strictly limited. Our model also implies that the surface of the cortical actin cap region can be tightly coated with unknown molecule(s) or use macromolecular crowding to exclude inappropriate molecules such as JUNO from the region (Kuznetsova et al., 2014). Identification of the novel molecules that cooperate with CD9 to exclude GPI-APs in the cortical actin cap region should help determine the importance of CD9 in fertilization.

CD81, a homologue of CD9, has a large intramembrane cavity for the uptake of cholesterol within its transmembrane region, and is kept in a closed (cholesterol-bound) conformation until formation of a complex with CD19, cis-partner of CD81 (Zimmerman et al., 2016). It is well known that the GPI-APs have a preference for gathering in a lipid raft, which consists of microdomains enriched in cholesterol and sphingolipids. Interestingly, cholesterol depletion using MβCD in mouse oocytes has been reported to show a decrease in fertilization rate (Buschiazzo et al., 2013). However, cholesterol is less likely to maintain appropriate surface compartmentalization of GPI-APs through the involvement of CD9, because the compartmentalization of GPI-APs in the cortical actin cap was not influenced by MβCD treatment (Fig. S3). Furthermore, a recently-published study provided supporting evidence that CD9 spontaneously undergoes transitions between the open and closed conformations without cholesterol involvement, unlike CD81 (Umeda et al., 2020).

It is noteworthy that JUNO, which had exuded into the cortical actin cap in CD9-deficient oocytes, was incapable of binding to the recombinant IZUMO1 (Figs 2D and 3C). There are two possibilities for this: (1) IZUMO1-JUNO binding in oocytes requires another molecule existing in the microvilli, such as a secondary IZUMO1 receptor as we had previously claimed (Inoue et al., 2015); or (2) IZUMO1 is unable to interact with JUNO because of the steric hindrance or embedding with other molecule(s) in the cortical actin cap region. The former possibility may be less likely because the IZUMO1-JUNO complex was co-crystallized (Aydin et al., 2016; Ohto et al., 2016). We therefore think that the latter possibility is more plausible. As chromosomes are confined underneath the cortical actin cap, the establishment of a dense coat on the outer leaflet may help physically protect the elaborate post-fertilization processes.

Therefore, this study shows that, in order to accomplish fertilization properly, tetraspanin CD9 contributes to multiple physiological functions including microvilli biogenesis, exosome vesiculation and a novel ‘molecular compartmentalization’ in mammalian oocytes (Fig. 6). It remains elusive how multiple molecular mechanisms coordinate to operate appropriate fertilization; however, we believe it indicates that, in general, CD9 plays an important role in cluster-induced curvature generation in microvilli or exosome biogenesis (Umeda et al., 2020). A new insight into molecular signalling and interaction thorough tetraspanin involvement would greatly advance understanding of the elaborate molecular mechanisms of gamete recognition and fusion.

B6D2F1 and ICR mice (Mus musculus) were purchased from Japan SLC. The B6D2F1 mouse strain was used as wild type in all experiments. IZUMO1 knockout (Inoue et al., 2005), IZUMO1-mCherry transgenic (Satouh et al., 2012) and Acr-EGFP (Nakanishi et al., 1999) mice were kindly provided by Osaka University. All animal studies were approved by the Animal Care and Use Committee of Fukushima Medical University, Japan, and performed under the guidelines and regulations of Fukushima Medical University.

An α-mouse IZUMO1 monoclonal antibody (Mab125) was conjugated to Alexa Fluor 647 (1:100). FITC-conjugated α-mouse CD9 monoclonal (MZ3, 1:1000, 124808), Alexa Fluor 647-labelled α-mouse JUNO monoclonal (TH6, 1:2000, 125110), Alexa Fluor 647-labelled α-mouse CD55 monoclonal (RIKO-3, 1:500, 131805) and unlabelled α-mouse JUNO (12A5, 1:500, 125002) monoclonal antibodies were purchased from BioLegend. Alexa Fluor 546- and 647-labelled α-mouse integrin α6 monoclonal (GoH3) antibodies were purchased from Santa Cruz Biotechnology (1:200, sc-19622AF546) and BioLegend (1:100, 313609), respectively. FITC-labelled α-mouse integrin β1 monoclonal (HMb1-1, 1:100, AB_2572449) and unlabelled α-mouse CD9 (EM-04, 1:1000, AB_11155997) monoclonal antibodies were purchased from Thermo Fisher Scientific. Unlabelled α-mouse GAPDH (5A12, 1:500, 016-25523) and α-mouse CD81 (Eat2, 1:200, MCA1846AA) monoclonal antibodies were purchased from FUJIFILM Wako Pure Chemical Corporation and Bio-Rad, respectively. Unlabelled α-mouse CD315 polyclonal antibody (1:100, AF4495) was purchased from R&D Systems. Recombinant IZUMO1-FC (1:150), α-mouse CD81 monoclonal and α-mouse CD315 polyclonal antibodies were detected by secondary Alexa Fluor 546-labelled α-mouse IgG (Thermo Fisher Scientific, 1:250, A11030), Alexa Fluor 647-labelled α-Armenian hamster IgG (Jackson ImmunoResearch, 1:250, 127-035-160) and Alexa Fluor 647-labelled α-Sheep IgG (Jackson ImmunoResearch, 1:250, 713-605-147), respectively.

Superovulation was induced in each female mouse (>8 weeks old) by injecting 7.5 IU of equine chorionic gonadotropin (eCG) and 7.5 IU of human chorionic gonadotropin (hCG) at 48 h intervals. Sixteen hours after hCG injection, oocytes were collected from the oviduct and placed in 50 μl of TYH medium. The zona pellucida was removed from the oocytes via treatment with 1.0 mg ml⁻¹ of collagenase (FUJIFILM Wako Pure Chemical Corporation) (Yamatoya et al., 2011). Zona-free oocytes were incubated with 0.5 µg ml⁻¹ MZ3-FITC and 0.25 µg ml⁻¹ TH6-Alexa647 with 1 μg ml⁻¹ of Hoechst 33342 (Thermo Fisher Scientific) for 1 h at 37°C in TYH medium. Mouse spermatozoa were collected from the cauda epididymis and capacitated in vitro for 1.5 h in a 200 µl drop of TYH medium with 0.5 µg ml⁻¹ Mab125 that was covered with mineral oil. Then, 5×10⁵ spermatozoa ml⁻¹ was inseminated in the oocytes for 15 min at 37°C with 5% CO₂. After the eggs were washed several times in TYH medium by transferring spots of oocytes containing media, the eggs were fixed with 4% paraformaldehyde and 0.5% polyvinylpyrrolidone (PVP) in PBS for 30 min at room temperature, then observed in glass-bottom dishes (No. 0, MatTek Corporation). A 100× oil-immersion objective (numeric aperture 1.49) was used to capture confocal images with an A1R microscope (Nikon). The pinhole was set at 3.0 airy unit. For 3D reconstruction, 100-120 fluorescent images were taken at 1 μm intervals on the z-axis, and then 3D images were reconstructed using the built-in software NIS-Elements ver. 4.3 (Nikon).

A pair of oligonucleotides (5′-CACCGTGAGATAGCCGCCGCCGTCT-3′ and 5′-AAACAGACGGCGGCGGCTATCTCAC-3′ for CD9; 5′-CACCGAGTGGTGGCAGATCCTCTTG-3′ and 5′-AAACCAAGAGGATCTGCCACCACTC-3′ for JUNO) as guide sequences were annealed and cloned into a BbsI cut pX330 vector (Addgene plasmid #42230). pX330 plasmid DNA was purified and diluted into 5 mM Tris-HCl (pH 7.4)/0.1 mM EDTA of a final 5 ng μl⁻¹ concentration. Plasmid DNA was injected into the pronuclei of fertilized eggs with glass capillaries attached to a micromanipulator (FemtoJet, Eppendorf). After microinjection, the zygotes were transferred and cultured in KSOM (ARK Resource) at 37°C with 5% CO₂, and two-cell embryos were transferred into pseudopregnant ICR female mice. F3 and F4 generation mice were used for the experiments, and all genotyping was performed by PCR on tail-tip DNA using CD9 5′-TGTGGGTATAGCCCCAGACG-3′ for wild-type or 5′-CATCCTTGTGGGTATAGCGG-3′ for KO and 5′-CCCTCATGATGCTGGTTGGT-3′ as forward and reverse primers, respectively, and JUNO 5′-CAAGAGGATCTGCCACCACT-3′ for wild-type or 5′-TAGGACTGCCCACAACCCTG-3′ for KO and 5′-AGATCCATTCCTTGCCTGGC-3′ as forward and reverse primers, respectively. CD9 and JUNO knockout mice have been deposited at the RIKEN BioResource Research Centre [RIKEN BRC (https://web.brc.riken.jp/en/); accession numbers RBRC10274 and RBRC10276, respectively].

Superovulated oocytes were washed three times using PBS, treated with SDS-sample buffer [62.5 mM Tris-HCl (pH 6.8), 2% SDS, 10% glycerol] and a protease inhibitor cocktail (FUJIFILM Wako Pure Chemical Corporation), then used for western blotting under non-reduced conditions. The membranes were probed with primary antibodies [1 μg ml⁻¹ α-CD9 (EM-04), 1 μg ml⁻¹ α-JUNO (12A5) and 1 μg ml⁻¹ α-GAPDH (5A12)] followed by secondary antibodies conjugated to HRP (Jackson ImmunoResearch, 1:5000, 115-036-062 and 112-006-143). Chemiluminescence reactions were performed with ECL Prime (Merck Millipore).

B6D2F1 mouse spermatozoa were collected from the cauda epididymis and were capacitated in vitro for 1.5 h in a 200 µl drop of TYH medium that was covered with mineral oil. Superovulation was induced in each female mouse (>8 weeks old) as described above. Oocytes were collected from the oviduct, placed in 100 μl of TYH medium and incubated with 2×10⁵ spermatozoa ml⁻¹ at 37°C with 5% CO₂. After 4 h insemination, the oocytes were washed with TYH medium and the fertilization rate was evaluated by pronuclear formation.

Zona-free oocytes were prepared as described above. For the actin depolymerization treatment, zona-free oocytes were incubated with 10 μM latrunculin B (Merck) or 0.2% DMSO in TYH medium for 1 h at 37°C in 5% CO₂ in advance (Runge et al., 2007). The oocytes were then incubated with 0.5 μg ml⁻¹ MZ3-FITC and 0.5 μg ml⁻¹ TH6-Alexa647, or 5 μg ml⁻¹ IZUMO1-FC or 1 μg ml⁻¹ GoH3-Alexa546 and 0.5 μg ml⁻¹ TH6-Alexa647 with 1 μg ml⁻¹ of Hoechst 33342 for 1 h at 37°C in 5% CO₂ in TYH medium. After being washed several times in TYH medium, the oocytes were placed in FHM medium (Merck Millipore) and then observed on glass-bottom dishes (No. 0, MatTek). The observation conditions were the same as those described above in ‘Fluorescence imaging for fixed eggs’ and the samples were kept at 37°C.

The IZUMO1-PA or -mCherry stably expressing COS-7 cells were treated with 5 mM EDTA-PBS at 37°C for 10 min, washed three times with PBS, centrifuged at 6000 g for 1 min and suspended in TYH medium (LSI Medience Corporation). Zona-free oocytes were incubated with the collected cells at 37°C in the TYH medium for 30 min. The oocytes were then washed and moved to drops of TYH medium, and the attached cells were counted under an inverted microscope (Olympus IX71). COS-7 cells were obtained from RIKEN BRC. This cell line has been authenticated as mycoplasma free by PCR.

The mature domain of mouse IZUMO1 (residues 22-319) cDNA without the secretion signal peptide region (residues 1-21) was expressed fusing with the CH2 and CH3 domains of the mouse IgG2a heavy chain and the hinge region from the expression vector pCMV6-AC-FC-S (OriGene). This construct uses a human interleukin 2 (IL2) leader sequence at the N terminus. For expression, 293T cells were transiently transfected with the IZUMO1-FC DNA using Polyethyleneimine ‘Max’ (Polysciences). These cells were maintained in 100 ml of DMEM supplemented with 2% fetal bovine serum, 1 mM sodium pyruvate, 1× non-essential amino acids (Nacali Tesque), 2 mM L-glutamine, 1000 units of penicillin and 1 mg of streptomycin at 37°C. The medium was collected after 4 days, and debris was filtrated using a 0.45 μm filter before loading on HiTrap rProtein A FF columns (GE Healthcare). The column was washed with 20 mM phosphate buffer (pH 7.0), eluted with 0.1 M sodium citrate (pH 3.0) and neutralized with 160 mM Tris-HCl (pH 9.0). Peak fractions were pooled and protein concentration measured using the Bradford protein assay. Finally, the protein purity was judged using SDS-PAGE.

Zona-free oocytes were prepared as described above, and placed on poly-L-lysine-coated cover glass for 30 s at room temperature. Primary fixation was performed in 4% paraformaldehyde, 2.5% glutaraldehyde, 2 mM CaCl₂ and 0.5% PVP in 250 mM HEPES buffer (pH 7.2). After fixation for 3 h at 4°C, the samples were washed with PBS three times and transferred into 2% osmium tetroxide in PBS at 4°C for 1 h. The samples were then dehydrated in an ethanol series and dried in liquid CO₂ using a critical point drying apparatus (JCPD-5; JEOL). Next, the samples were transferred onto a stub with silver paste plus thinner and coated with osmium by an HPC-1SW coater (Shinku Device). An SU8220 SEM (Hitachi) was used for imaging according to the manufacturer's protocol.

Zona-free oocytes were fixed in 4% paraformaldehyde and 0.5% PVP in PBS for 30 min, then permeabilized with 0.25% Triton X-100 and 0.5% PVP in PBS for 20 min. After washing with FHM medium three times, the oocytes were incubated in FHM medium with 5 units ml⁻¹ BODIPY 558/568 phalloidin (Thermo Fisher Scientific) and 1 μg ml⁻¹ Hoechst 33342 for 10 min, followed by washing three times in FHM medium. The oocytes were then placed in FHM medium on glass-bottom dishes (No. 0) and imaged using a Nikon A1R confocal microscope equipped with 405- and 561-nm laser lines using a 100× oil-immersion objective lens.

Zona-free oocytes were incubated with the appropriate antibodies and 1 μg ml⁻¹ of Hoechst 33342 in TYH medium for 90 min, then treated with 2 mM MβCD (Sigma-Aldrich) in FHM medium for 30 min at 37°C to deplete cholesterol (Buschiazzo et al., 2013). After washing with FHM medium three times, the oocytes were placed in FHM medium on glass-bottom dishes (No. 0) and imaged using a Nikon A1R confocal microscope equipped with a 100× oil-immersion objective at 37°C.

FRAP analysis of the oocytes was carried out as previously described (Nagaya et al., 2008). After immunostaining with 0.5 µg ml⁻¹ TH6-Alexa647 (JUNO) (BioLegend, 1:1000, 125110) or 1 µg ml⁻¹ RIKO-3-Alexa647 (CD55) (BioLegend, 1:500, 131805) for 2 h, the oocytes were washed three times in TYH medium and then placed in FHM medium on glass-bottom dishes (No. 0). The oocytes were kept at 37°C. A spot in the actin cap region was photobleached with a 640-nm laser from the Nikon A1R confocal microscope through a 60× water-immersion objective lens (numeric aperture 1.27) at 0.026 mW, and the recovery of the fluorescence was monitored at 2 min intervals. Fluorescence intensity before photobleaching was used to normalize the intensity during recovery.

Spermatozoa from IZUMO1-mCherry and Acr-EGFP double transgenic mice (Izumo1 null genetic background) (Satouh et al., 2012) were collected from the cauda epididymis, and were capacitated in vitro for 2 h in a 200 µl drop of TYH medium that was covered with mineral oil. Zona-free oocytes were prepared as described above and incubated with 1 μg ml⁻¹ α-mouse integrin α6 antibody (GoH3-Alexa647) (BioLegend, 1:100, 313609) for 90 min at 37°C in 5% CO₂ in TYH medium. After being washed with TYH medium three times, the oocytes were incubated with 5×10⁵ spermatozoa ml⁻¹ for 10 min at 37°C with 5% CO₂. The eggs were fixed with 4% paraformaldehyde and 0.5% PVP in PBS for 30 min at room temperature, then placed in FHM medium on glass-bottom dishes (No. 0) and imaged using a Nikon A1R confocal microscope equipped with a 60× water-immersion objective lens. The numbers of bound spermatozoa were counted.

All image and statistical analyses were performed using the open source software Fiji (Schindelin et al., 2012) and OriginPro 2020 (Light Stone), respectively.

We thank Drs Yuhkoh Satouh and Ken Sato (Gunma University) for thoughtful discussions.