A Novel Live Imaging-Based Assay for Visualising Species-Specic Interactions in Gametic Adhesion Molecules

Successful gametic fusion requires species-specic membrane adhesion. However, the interaction of adhesion molecules in gametes is dicult to study in real time through low-throughput microscopic observation. Therefore, we developed a novel live imaging-based adhesion molecule (LIAM) assay to study gametic adhesion molecule interactions in cultured cells. First, we modied a fusion assay previously established for fusogens introduced into cultured cells, and conrmed that our live imaging technique could visualise cell-to-cell fusion in the modied fusion assay. Next, instead of fusogen, we introduced adhesion molecules including a mammalian gametic adhesion molecule pair, IZUMO and JUNO, and detected their temporal accumulation at the contact interfaces of adjacent cells. Accumulated IZUMO or JUNO was translocated to the opposite cells; the mutation in amino acids required for their interaction impaired accumulation and translocation. By using the novel LIAM assay, we investigated the species specicity of IZUMO and JUNO of mouse, human, hamster, and pig in all combinations. IZUMO and JUNO accumulation and translocation were observed in conspecic, and some interspecic, combinations, suggesting potentially interchangeable combinations of IZUMO and JUNO from different species.


Introduction
Sexually reproducing organisms commonly reproduce via the union of different sex cells. Fertilisation is the nal process of sexual reproduction; it is achieved through the recognition, adhesion, and fusion of gametes in processes regulated by fertilisation molecules on gametes. Fertilisation molecules have been identi ed in several species. For example, in mammals, sperm that have passed through the zona pellucida approach the egg and ultimately fertilise it through interactions of fertilisation molecules, such as IZUMO on the sperm and CD9 and JUNO on the egg 1-3 . Mutants of IZUMO, CD9, and JUNO are completely infertile. Co-crystal structure analysis has demonstrated that IZUMO and JUNO interact as adhesion molecules, and that species-speci c binding depends on the conserved amino acid sequence, e.g., W148A of IZUMO and W62A of JUNO, on the interaction surface 4,5 , although the machinery required for membrane fusion during fertilisation remains unclear. To date, only IZUMO and JUNO have been identi ed as a gametic adhesion molecule pair. A recent study also reported that the sperm proteins SPACA6, TMEM95, SOF1, FIMP are required for mammalian fertilisation 6-9 .
In owering plants, two immotile sperm cells are conveyed by the pollen tube to the female gametes, and the adhesion molecule GAMETE EXPRESSED 2 (GEX2) and fusogen GENERATIVE CELL-SPECIFIC 1/HAPLESS 2 (GCS1/HAP2) on the sperm cell membrane promote fertilisation with the two female gametes, i.e. the egg cell and central cell [10][11][12] , in a unique fertilisation process referred to as double fertilisation. GEX2 has a lamin-like domain similar to the Ig-like domain contained in IZUMO. The green alga Chlamydomonas reinhardtii has no GEX2 orthologue; however, FUS1, which contains a lamin-like domain, is expressed on plus-type cells 13 . Recently, FUS1 partner molecule, Minus Adhesion Receptor 1 (MAR1), was identi ed, which was required not only for gamete adhesion but for focused localization of GCS1/HAP2 at the mating structure 14 . GCS1/HAP2 is an ancestral eukaryotic fusogen widely conserved in plants and protists 15 . Crystal structure studies have revealed that GCS1/HAP2 shows homology with the Caenorhabditis elegans somatic cell fusogen Epithelial Fusion Failure 1 (EFF-1) 16 and class II viral fusion proteins (e.g. Dengue, Zika, and West Nile viruses) 17 . Whether the GCS1/HAP2 fusion mechanism is unilateral or bilateral remains controversial 18, 19 . GCS1/HAP2 does not exist in mammals, and fusogens have not been identi ed in mammalian gametes. Current studies are seeking to identify the complete set of fertilisation molecules in several species. Missing partner molecules and the function and action mechanisms of known fertilisation molecules require investigation to understand the precise molecular mechanisms of fertilisation.
Genetic loss-of-function experiments with phenotypic analysis have revealed that fertilisation-related molecules are required for gamete adhesion or fusion. However, the precise molecular interactions driving cell membrane adhesion and fusion remain to be elucidated. Recent gain-of-function experiments have investigated the molecular functions of membrane proteins. For example, nematode speci c proteins EFF-1 and AFF-1 have been shown to fuse nematode, insect, mammalian and viral membranes 16,20−23 .
More recently, the muscle-speci c protein Myomaker was transfected with Myomerger (also called Myomixer and Minion) into broblasts to investigate their biochemical function in cell-membrane fusion 24 . Molecular adhesiveness of a gametic adhesion molecule pair IZUMO-JUNO has been shown by ectopic expression of IZUMO or JUNO in cultured somatic cells 4,25,26 . Gain-of-function experiments in heterologous cell systems are powerful tools for the elucidation of molecular functions with characteristic activities. Combining this technique with live-cell imaging would allow us to visualise molecular dynamics and functions in real time.
Species-speci c and -preferential interactions among fertilisation molecules also require further investigation. Species speci city has been studied in mammals for several decades. Ovulated eggs are surrounded by the zona pellucida; Yanagimachi et al. found that golden hamster zona-free oocytes could be fertilised with sperm of various species 27 . Some proteins involved in male-female interactions have been suggested to be involved in species-speci c interactions. The zona pellucida glycoprotein ZP2 is required for species-speci c sperm-egg binding through the zona pellucida in mice and humans 28 . The glycosylphosphatidylinositol (GPI)-anchored protein Bouncer is required for species-speci c sperm entry into the egg, and enables cross-species fertilisation between medaka and zebra sh through gene swapping 29 . Extracellular protein interactions were exploited in the AVidity-based Extracellular Interaction Screening (AVEXIS) assay, demonstrating conspeci c interaction between IZUMO and JUNO in mouse cells; the results suggested that hamster JUNO can interact with human, mouse, and pig IZUMO 30 , consistent with the crossing result obtained by Yanagimachi et al 31 . However, in owering plants, gametic interaction appears to be less critical for species-speci c fertilisation than the progamic phase of pollenpistil interactions [32][33][34] .
To elucidate the molecular dynamics and function of molecules involved in gametic interactions such as gamete adhesion and fusion, we developed a system to analyse the functions of fertilisation molecules through live-cell imaging in cultured animal cells. For this gain-of-function approach, we modi ed a fusion assay developed by Valansi et al. 18 and transfected cultured baby hamster kidney (BHK) cells to express a gametic fusogen GCS1/HAP2, which promoted their cell-to-cell fusion. We con rmed that cellto-cell fusion of BHK cells expressing fusogens could be captured by our live-cell imaging technique using a disc scanning confocal system embedded within the incubator. Then, we introduced adhesion molecules to explore their effects in cultured cells, which lead us to develop a live imaging-based adhesion molecule (LIAM) assay. LIAM assay was shown to be useful for the screen of speciesspeci city in gametic adhesion molecules IZUMO and JUNO.

Results
Fusogen promotes cell-to-cell fusion in our live-cell imaging system We aimed to develop a system to study fertilisation molecules in animal cultured cells by live imaging. Various cultured animal cell lines, such as COS-7 and HEK293T, have been used for transfection and expression in previous fertilisation molecule studies; for example, molecular adhesiveness between IZUMO and JUNO was shown by mixing gametes and transfected cultured-cells 3,26,35 . In this study, because we focused on cell migration and morphology for frequent contact of cells under microscopy, we selected a line of motile BHK cells that did not spontaneously adhere or fuse with each other (Fig. S1).
BHK cells were also used in the original fusion assay 18 . We investigated whether cell fusion activity could be detected in known fertilisation molecules using our live-cell imaging technique by transfecting BHK cells with RFPcyto as a negative control, and GCS1/HAP2 to evaluate multinucleation rates. Ectopic expression of fusogens sometimes indicates toxicity to cells or the whole organism 16,18,21,36 ; therefore, tight regulation was required to ensure expression at the preferred times. For this purpose, we used a mifepristone-inducible system as used in the fusion assay 18 .
To capture cell-to-cell fusion by live imaging, we used a Nipkow disc scanning confocal system embedded within the incubator to monitor cells in 1 mm × 1 mm areas every 6 min under stable lowphototoxicity conditions (Fig. 1). In each captured video, 130-476 cells expressing RFPcyto were monitored. When the gametic fusogen GCS1 (Arabidopsis thaliana GCS1; AtGCS1) was expressed, cell-tocell fusion was observed (Fig. 1A, arrows; Video 1). At 12 h after induction, the multinucleation rate increased signi cantly, to 2.2 ± 0.4% (n = 4; P < 0.05; Dunnett's test; Fig. 1B). To con rm that cell-to-cell fusion depended on the fusogen activity of the expressed proteins, we transfected the loop-deletion GCS1 (GCS1∆loop). Amino acids 166-178 in AtGCS1 represent a hydrophobic loop structure, which is necessary for membrane fusion 17 . The multinucleation rate was 1.2 ± 0.2% (n = 3), which was not signi cantly different from the negative control (Fig. 1B). This result suggested that fusogen-dependent cell-to-cell fusion was visualised by live imaging. Our live imaging system is useful to monitor expression of a uorescent marker upon cell-to-cell fusion. The dependency of individual fusion on the expression of transfected genes from 4 to 12 h after expression induction is shown in Fig. 1C. In experiments for wild-type GCS1, a combination of two cells labelled with red uorescent protein (RFP) was more likely to fuse than a combination of labelled and non-labelled cells. This is partly due to the frequent fusion of daughter cells after cell division. Our result is consistent with the previous report that GCS1, as well as EFF-1, bilaterally promoted cell-to-cell fusion in the fusion assay 18 . Together, these results suggest that our live imaging technique is compatible with a modi ed fusion assay for the study of fusogen functions in cultured cells.
Adhesion molecules accumulate at the contact interfaces of adjacent cells The fusion assay has never previously been used to evaluate gametic adhesion molecule function; conventional methods include adhesion and aggregation assays 37,38 . To develop a simple method suitable for live imaging-based studies, we introduced adhesion molecule pairs into the modi ed fusion assay.

First, we transfected the somatic adhesion molecule E-cadherin into BHK cells. Cadherins have been
shown to play a role in Ca 2+ -dependent cell-cell adhesion 39 . The transfection of a single cadherin gene can potentially induce interaction of cadherin expressed in different cells (homophilic adhesion), eliminating the necessity of introducing an adhesion molecule pair such as IZUMO and JUNO (heterophilic adhesion). Cadherin was visualised by translational fusion with GFP, and no toxicity was observed despite the lack of an inducible gene expression system for cadherin in this experiment. Ecadherin-expressing BHK cells in our system did not adhere to neighbouring cells to show aggregation. Ecadherin appeared to be localised at the cell membrane and secretory pathways of the cell, showing no polarisation ( Fig. 2A). However, upon temporary contact between E-cadherin-expressing cells, E-cadherin tended to accumulate at the contact interface ( Fig Next, we introduced mouse IZUMO and JUNO to determine whether a similar temporal accumulation at the contact interface of adjacent cells might occur even in heterophilic adhesion molecules. We developed a mixing assay, in which cells expressing IZUMO and JUNO were mixed at 4 hours after independent transfection and were incubated for 20 hours. IZUMO and JUNO were designed as translational fusion with uorescent proteins. At 4 to 12 hours after the induction of IZUMO and JUNO expression by mifepristone, we found that IZUMO temporally accumulated at the contact interfaces of We examined whether IZUMO accumulation is dependent on the interaction of IZUMO and JUNO expressed in adjacent cells. When IZUMO and JUNO were expressed simultaneously in the same cell by co-transfection, neither molecule accumulated on the cell surface (0%, n = 63; Fig. 2E). Next, we constructed a mutant IZUMO lacking the β-hairpin region (IZUMO∆) and examined the interaction between IZUMO∆ and JUNO. The central β-hairpin region of IZUMO is required for JUNO binding, as has been shown by co-crystal structure analysis 4,5 . When IZUMO∆-and JUNO-expressing cells were mixed, accumulation of IZUMO was rarely observed (7% of cases, n = 59; Fig. 2E). We also co-transfected JUNO with CD9, an egg tetraspanin required for gamete fusion 2 , to determine whether CD9 modulates IZUMO accumulation. When IZUMO-and JUNO/CD9-expressing cells were mixed, IZUMO accumulation and frequency were unchanged, consistent with the nding that CD9 is independent of gamete adhesion 2,40 (50%, n = 42; Fig. 2E). These results demonstrate the usefulness of this assay for visualising the dependence of adhesion molecule interactions on the a nity of introduced adhesion molecule pairs. We named this assay system the live imaging-based adhesion molecule (LIAM) assay, which we propose for the examination of transfected adhesion molecules at cell contact sites.
Accumulation and translocation of adhesion molecules in the LIAM assay depended on critical amino acids Subsequent experiments were conducted using high-throughput analysis in a non-inducible gene expression system (cytomegalovirus promoter), because we found that the inducible gene expression system was unnecessary for cadherin, IZUMO, and JUNO (translational fusion with uorescent proteins). Prior to contact between mouse IZUMO-and JUNO-expressing cells, these adhesion molecules were distributed evenly throughout the cell, as also occurs in the inducible gene expression system.
Accumulation was observed upon cell contact; however, both IZUMO and JUNO showed accumulation To determine whether accumulation was dependent on the amino acid residues of IZUMO and JUNO for molecular interaction, we performed a LIAM assay using mutant IZUMO and JUNO, according to previous crystal structure analysis results 4,5 (Fig. 3A). When mouse mutant IZUMO W148A-expressing cells were mixed with mouse JUNO-expressing cells, accumulation was not observed (0% of 45 contacting cells; Fig.  3B). When mouse IZUMO-expressing cells and mouse mutant JUNO W62A-expressing cells were mixed, accumulation was again not observed (0% of 67 contacting cells; Fig. 3B). Prior to cell contact, IZUMO and JUNO expression and localisation were unchanged by these point mutations (Fig. 3A). Next, we tested the sperm transmembrane protein SPACA6, which is required for fertilisation 6,7 , instead of IZUMO, which shows structural similarity with SPACA6 6 (Fig. S2), and found that neither SPACA6 nor JUNO accumulated (Fig. 3A). These results indicate that IZUMO and JUNO accumulation depends on amino acids critical for their interaction, according to the LIAM assay.
In 38% (15/39) cells showing IZUMO accumulation, IZUMO was found to translocate to the plasma membrane of the opposite cell (Fig. 3C). Several uorescent foci of IZUMO or JUNO on the opposite cell were observed upon the detachment of contacting cells (Video 5). IZUMO was translocated to JUNOexpressing cells in 27% of contacting cells (15/54; Fig. 3D) and JUNO was translocated to IZUMOexpressing cells in 7.4% of contacting cells (4/54; Fig. 3D). This translocation of IZUMO and JUNO was unilateral, not bilateral, in each cell pair, and translocation was selectively observed for either IZUMO or JUNO in each cell pair. In addition, IZUMO or JUNO translocation was always observed after IZUMO accumulation, but not necessarily after JUNO accumulation, perhaps due to the di culty associated with visualising JUNO accumulation on the cell membrane.
This translocation was dependent on the amino acid sequence of IZUMO and JUNO required for the interaction (Fig. 3D). When mutant IZUMO W148A or mutant JUNO W62A was used, neither translocation not accumulation was observed (Fig. 3D). SPACA6 and JUNO also did not induce translocation (Fig. 3D).
In subsequent experiments, this translocation of adhesion molecules was also monitored in the LIAM assay.

Species speci city in JUNO and IZUMO interactions suggested by the LIAM assay
Using the LIAM assay, we investigated the species speci city of IZUMO and JUNO from mouse, human, hamster, and pig ( Fig. 4; Tables 1-4). Among these four mammalian species, mice are most closely related to hamsters, and humans and pigs are phylogenetically close to each other 41 . Molecular phylogenetic analysis results for IZUMO and JUNO proteins were consistent with the phylogenetic relationships among these four species (Figs. S3-5). The mouse IZUMO protein shares 61%, 52%, and 50% homology (identity) with hamster, human, and pig IZUMO proteins, respectively (Figs. S3, S5). In contrast, the mouse JUNO protein shares 74%, 68%, and 65% homology with hamster, human, and pig JUNO proteins, respectively (Figs. S4, S5). These four species were selected because species speci city has been observed biochemically in their IZUMO and JUNO interactions, in some but not all combinations, using the AVEXIS assay 3 . In the AVEXIS assay, IZUMO and JUNO interaction was shown in all conspeci c combinations, hamster JUNO and all other three species, and mouse JUNO and human IZUMO but not in human JUNO and mouse IZUMO. We examined all combinations of IZUMO and JUNO from mouse, human, hamster, and pig cells using the LIAM assay. We began with conspeci c combinations. Mouse IZUMO and JUNO showed accumulation in 72% (39/54; Fig. 3; Table 1) and 37% (20/54; Fig. 3; Table 2) of contacting cells, respectively. Translocation of mouse IZUMO and JUNO was also observed, in 27% (15/54; Table 3; Fig. 3C) and 7.4% (4/54; Table 4) of contacting cells, respectively. This nding is consistent with previous AVEXIS assay results 3 , which showed that mouse IZUMO and JUNO show relatively strong binding. In this study, we observed IZUMO accumulation as often as JUNO accumulation in human IZUMO × human JUNO interactions (IZUMO accumulation: 28%, 14/51, Table 1; JUNO accumulation: 29%, 15/51, Table 2). In this combination, IZUMO translocation was not observed (Table 3), whereas JUNO translocation was frequently observed (64%, 33/51, Table 4, Fig. 4A). In hamster IZUMO × hamster JUNO, IZUMO and JUNO accumulation (IZUMO: 77%, 41/57, Table 1; JUNO: 28%, 16/57, Table 2) and translocation (IZUMO: 31%, 18/57, Table 3; JUNO: 28%, 16/57, Table 4, Fig. 4B) were observed, consistent with mouse IZUMO and JUNO. In pig IZUMO × pig JUNO interactions, we observed IZUMO accumulation at a low rate (24%, 10/41, Table 1, Fig. 4C). Conspeci c interaction of IZUMO and JUNO was con rmed for all combinations, and the frequencies of IZUMO and JUNO accumulation and translocation differed among species.    (41) Next, we examined heterospeci c interaction between human IZUMO and mouse, hamster, and pig JUNO. Neither accumulation nor translocation was detected in any combination (Tables 1-4). When hamster IZUMO-expressing cells and mouse JUNO-expressing cells were mixed, IZUMO and JUNO accumulation were detected, although at lower rates than for conspeci c combinations (mouse IZUMO × mouse JUNO, hamster IZUMO × hamster JUNO) (Tables 1 and 2). This result is consistent with the crossing results reported by Yanagimachi et al. 27 and those of a previous AVEXIS assay 3 . No translocation was observed (Tables 3 and 4). Finally, we combined pig IZUMO-expressing cells with mouse, human, and hamster JUNO-expressing cells. When pig IZUMO-expressing cells and mouse JUNO-expressing cells were mixed, neither accumulation nor translocation was observed (Tables 1-4). However, when pig IZUMO-expressing cells and human JUNO-expressing cells were mixed, IZUMO and JUNO accumulation was detected. Unexpectedly, translocation of both IZUMO and JUNO was detected in this heterospeci c combination of pig IZUMO and human JUNO (Tables 1-4, Fig. 4D). This strongly positive result was obtained using our LIAM assay, but has not been tested previously using other assays.

Discussion
In this study, we developed the LIAM assay to study gametic adhesion molecules in cultured animal cells.
The LIAM assay represents a powerful complement to the fusion assay 18 for microscopic examination of the functions of fertilisation molecules in various organisms under highly controllable conditions. As previously observed in fusogens including plant GCS1/HAP2 and nematode EFF-1, the functions of the mammalian adhesion molecule pair IZUMO and JUNO were assessed at the contact interfaces of adjacent cells. Two phenotypes of adhesion molecule dynamics, i.e., accumulation and translocation, were detected by the LIAM assay (Fig. 3). The accumulation of homophilic and heterophilic somatic adhesion molecules 39,42,43 , as well as gametic IZUMO and JUNO 3 , has been reported at the interface of adhered cultured cells; however, our live imaging-based assay suggested that temporal contact was su cient to induce the intracellular accumulation of adhesion molecules in non-aggregating cells. Such translocation is an interesting phenomenon that may be caused by protein extraction from the membrane of the adjacent cell during contact. Monodirectional translocation (i.e. selective translocation of IZUMO or JUNO in a single cell pair) and stable foci of translocated proteins suggest that some cell activity, such as endocytosis, trogocytosis 44 , or exosome receprion 45 within a cell pair, is involved in the translocation and compartmentalisation of adhesion molecules, depending on the molecule combination.
Accumulation and translocation were strictly dependent on the structures of the tested molecules (Fig. 3). Single amino acid mutations in IZUMO or JUNO, which are critical for their interaction, abolished these signals completely. Other loop and domain deletions, as well as the swapping of IZUMO and SPACA6, also resulted in drastic decreases in these signals (Figs. 2 and 3). These results suggest that the LIAM assay successfully detected interactions within the adhesion molecule pairs. Consistently, IZUMO and JUNO that had simultaneously accumulated within a cell pair were localised together across the contact site (Fig. 3). The frequencies of accumulation and translocation detected by the LIAM assay were strongly correlated with conspeci c combinations in which direct interaction of IZUMO and JUNO have been shown biochemically using an AVEXIS assay 3 (Tables 1-4). This was also consistent with the detection of some heterospeci c but compatible combinations, such as mouse sperm IZUMO and human egg JUNO 25,26 or hamster egg JUNO 3 ( Table 1). The application of other analysis techniques, such as uorescence resonance energy transfer (FRET), would provide insights into the interaction of adhesion molecules observed in our LIAM assay; combining the LIAM assay with FRET would be a very powerful method for visualising molecular interactions in real time, which has not been attempted for gametic interactions in any organisms to date.
The advantages of the LIAM assay include higher throughput to identify potential molecule pairs showing heterophilic and homophilic adhesion activity among many combinations. After con rming conspeci c interaction of IZUMO and JUNO from mouse, human, hamster, and pig, we identi ed a candidate combination for strong heterospeci c binding: pig IZUMO and human JUNO. Among the tested four species, human and pig possess relatively close IZUMO and JUNO (Fig. S5). Evaluation of the adhesiveness of pig sperm and a human egg would be of interest, and to our knowledge has not been reported. The LIAM assay showed positive results for hamster JUNO with a wide range of IZUMO from mouse, pig, and hamster, but not human, which is largely consistent with previous reports that hamster zona-free eggs can be fertilised by sperm of various mammals 31 , and that hamster JUNO can interact with IZUMO from various species 30 . It may be possible to determine why hamster JUNO is compatible with IZUMO from various mammals in future studies, via detailed amino acid swapping analyses performed using the LIAM assay.
An anticipated direction of in vitro studies using LIAM and fusion assays is reconstitution of the gametic fertilisation machinery in cultured cells, which would be a powerful tool for the study of fertilisation molecules as a machinery complex. In LIAM assay, we did not observe cell-to-cell fusion, consistent with the previous report that ectopic expression of adhesion molecules is not su cient for cell-to-cell fusion of cultured somatic cells 46 . Reconstitution of the fertilisation machinery complex under microscopy is a promising direction for future research to clarify these molecular dynamics and functions. In mammals, partner molecules for the sperm proteins SPACA6, TMEM95, SOF1, FIMP, as well as a fusogen, remain elusive 6-9 . The identi cation of a complete set of fertilisation molecules, which is in progress in some organisms including Chlamydomonas 14 , would contribute signi cantly to reconstitution in cultured cells or liposomes.

Plasmid construction
Arabidopsis thaliana GCS1/HAP2 coding sequence (CDS) fragments were ampli ed from pSN30 (a gift from Dr. Shiori Nagahara) by polymerase chain reaction (PCR) using the primers listed in Supplemental

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.