A pharmacological study of Arabidopsis cell fusion between the persistent synergid and endosperm

Cell fusion plays important roles in various developmental events such as skeletal muscle formation, placenta development and fertilization in animals (Abmayr and Pavlath, 2012; Mayhew, 2014; Friedmann, 1962). By contrast, cell fusion is observed only three times during the life cycle in Arabidopsis thaliana (Maruyama et al., 2016). Interestingly, these three events occur during sexual reproduction. Plants produce two female gametes: the egg cell and the central cell, which are neighbors of two synergid cells. The synergid cells secrete pollen tube-attracting peptides to support the directional growth of the pollen tube, which carries the two sperm cells. After degeneration of one synergid cell and the concomitant release of sperm cells from the pollen tube, each sperm cell independently fuses with the egg cell or the central cell, which gives rise to the embryo or multinucleated endosperm, respectively (Yadegari and Drews, 2004; Berger et al., 2008). Within a few hours after the two cell fusions that conclude the double fertilization, the persistent synergid cell fuses with the endosperm. This phenomenon, designated as synergid–endosperm fusion (SE fusion), causes a dilution of the cytoplasm of the persistent synergid and prevents attraction of pollen tubes (Maruyama et al., 2015; Maruyama and Higashiyama, 2016).

At least two proteins are considered to have pivotal roles in plasma membrane fusion in flowering plants. GENERATIVE CELL SPECIFIC 1 [GCS1; also known as HAPLESS2, HAP2] is a sperm-specific membrane protein required for double fertilization (Mori et al., 2006; von Besser et al., 2006). GCS1/HAP2 belongs to the fusexin family of membrane fusion-accelerating proteins, such as Dengue virus E glycoprotein and C. elegans EFF-1 (Podbilewicz, 2014; Valansi et al., 2017; Pinello et al., 2017; Fédry et al., 2017). GAMETE EXPRESSED 2 (GEX2) is a sperm-specific membrane protein that participates in fertilization, probably through adhesion of male and female gametes (Mori et al., 2014). In contrast to double fertilization, the mechanism involved in SE fusion remains unknown.

Cell fusion involves not only specific proteins but also relies on common cellular components. For example, characteristic actin foci appear at the contact site of fusing myoblasts (Abmayr and Pavlath, 2012; Rochlin et al., 2010), the mating structure of fission yeast (Dudin et al., 2015) and green algae (Wilson et al., 1997), and at the acrosomes of spermatozoon (Tilney et al., 1973), indicative of a ubiquitous role of filamentous actin (F-actin) in cell fusions. In budding yeast, mating efficiency is reduced in temperature-sensitive mutants that are defective in the secretory pathway, a basic cellular mechanism widely conserved among eukaryotes (Grote, 2010). Here, we investigated the impact of major cellular activities on SE fusion. To our surprise, neither disruption of actin polymerization nor protein secretion abolished SE fusion. However, mitosis-affecting inhibitors severely impaired SE fusion. These findings imply the existence of unique mechanistic features in SE fusion.

To examine the role of F-actin formation in SE fusion, we used a pFWA::Lifeact-Venus transgenic plant. This line expresses a fluorescent marker for F-actin, Lifeact-Venus, in the central cell and endosperm (Kawashima et al., 2014; Riedl et al., 2008). Pistils from pFWA::Lifeact-Venus were pollinated with the pRPS5A::H2B-tdTomato nuclear marker line, and the ovules were analyzed at ∼8 h after pollination (hap) by performing confocal microscopy. When double-fertilization was successful, the tdTomato signal from sperm nuclei diffused in the zygote and endosperm nuclei. During a 6.5-h observation, the primary endosperm nucleus divided once or twice, and 69% of ovules showed migration of the endosperm Lifeact-Venus signal into the persistent synergid, indicative of SE fusion (Fig. 1A; Movie 1). When Drosophila myoblasts fuse, F-actin foci are observed at the contact sites between the two cell membranes prior to cell fusion (Sens et al., 2010). In contrast, during SE fusion, we did not observe any noticeable accumulation of F-actin between the endosperm and persistent synergid during the 10-min interval time-lapse imaging (Fig. 1A; Movie 1). In the fertilized ovules, treatment with the actin polymerization inhibitor latrunculin A abolished F-actin movement and induced destruction of actin cables (Fig. 1B; Movie 1) as reported previously (Kawashima et al., 2014). However, 55% of ovules showed Lifeact-Venus migration into the persistent synergid, indicating that this treatment did not prevent SE fusion.

To further investigate the involvement of F-actin in SE fusion, we monitored the fusion process in the pFWA::Lifeact-Venus plants that also expressed a dominant-negative form of ACTIN 8 (DN-ACTIN) from the central cell- and endosperm-specific FWA promoter (Kawashima et al., 2014). Aniline Blue staining at 2 days after pollination demonstrated that there was normal pollen tube attraction in the dominant-negative lines (Table S1). However, the dominant-negative lines exhibited retardation of endosperm nuclear proliferation (Table S2; Fig. 1D). This fertilization defect is caused by a loss of actin cables necessary for migration of the sperm nucleus toward the central cell nucleus after the plasma membrane fusion of the male and female gametes (Kawashima et al., 2014). Indeed, we confirmed the fragmentation of actin cables in those aberrant endosperms at 14 hap (Fig. 1D). Nevertheless, the signal of the Lifeact-Venus was detected in the persistent synergid of ∼80% of ovules (Fig. 1D,E). The frequencies of SE fusion in the DN-ACTIN lines were comparable with that in a control line that expressed wild-type ACTIN 8 as well as Lifeact-Venus from the FWA promoter (Fig. 1C,E; Kawashima and Berger, 2015). These results indicate that the mechanism of SE fusion is independent of F-actin dynamics.

Cell fusion-related proteins including fusogens, adhesion factors and cell wall-remodeling enzymes must be located in the plasma membrane or extracellular space. Most newly synthesized secretory proteins are transported to their destination via organelles that belong to the secretory pathway (e.g. the endoplasmic reticulum and Golgi). Therefore, late events of cell fusion are rapidly stalled by disruption of the secretory pathway during mating in budding yeast (Grote, 2010). To examine whether Arabidopsis SE fusion depends on the protein secretory pathway, the post-fertilization synergid and endosperm were monitored after treatment with brefeldin A (BFA). BFA targets guanine nucleotide exchange factors for the ADP-ribosylation factor that controls retrograde protein transport from the Golgi to the endoplasmic reticulum and severely disrupts protein secretion (Balch et al., 1992; Klausner et al., 1992). We confirmed that secretory-type CLOVER expressed from the central cell-specific promoter accumulated in the endosperm after BFA treatment (Fig. S1). After fertilization of ovules from a transgenic line expressing FWA-GFP, we monitored the influx of FWA-driven GFP into the persistent synergid cell in the presence or absence of BFA. Interestingly, the BFA treatment did not reduce the percentage of SE fusion (Fig. 2A,B; Movie 2). Therefore, the Arabidopsis ovule might accumulate sufficient levels of the SE fusion-related proteins on the plasma membrane prior to the first endosperm nuclear proliferation.

The disruption of protein production is also expected to block the supply of SE fusion-related factors to the plasma membrane. We performed time-lapse imaging of fertilized pFWA::FWA-GFP ovules in the presence or absence of translational or transcriptional inhibitors. If SE fusion-related factors were synthesized before fertilization, these ovules would have the potential to carry out SE fusion. However, translation inhibition by cycloheximide treatment strikingly decreased the percentage of SE fusion (Fig. 3B; Movie 3; to 8%), in contrast to the control experiment (Fig. 3A; 87%). A similar result was obtained from treatment with the transcriptional inhibitor cordycepin (Fig. 3C; Movie 4; 16%). These findings were unlikely to have resulted from the toxicity of these inhibitors because dynamic protoplasmic streaming was still observed during the experiments. We conclude that expression of new proteins after fertilization is required for SE fusion.

Knockdown of the main components of RNA polymerase II results in severe retardation of endosperm development (Pillot et al., 2010). Consistent with this, translational or transcriptional inhibition abolished nuclear proliferation in the endosperm as well as the induction of SE fusion (Fig. 3; Movies 3 and 4), which may imply a causal relationship between SE fusion and mitosis. This idea is also supported by observations that SE fusion often occurs during the first endosperm nuclear proliferation in wild-type ovules (Maruyama et al., 2015) and autonomously developing ovules (Motomura et al., 2016).

To examine the possible link between SE fusion and mitosis, we analyzed SE fusion under mitotic disruption induced by oryzalin, an inhibitor of microtubule polymerization. Pistils from a transgenic line expressing tagRFP-labeled TUBULIN ALPHA-5 by the endosperm-specific AGL80 promoter (pAGL80::tagRFP-TUA5) were manually pollinated with their own pollen. Then, ovules were incubated with or without oryzalin and subjected to time-lapse imaging. In the control, microtubule bundles formed an aster-like pattern around endosperm nuclei. The signal of tagRFP-TUA5 accumulated on the spindle during M-phase, and concomitantly, an influx of the tagRFP-TUA5 into the persistent synergid was observed in 87% of ovules (Fig. 4A; Movie 5). In the presence of oryzalin, the mitotic spindle did not form properly, and the nuclear division failed in endosperm. However, migration of tagRFP-TUA5 to the persistent synergid was still observed in 93% of ovules, indicating that microtubules and their role in mitosis are not required in SE fusion (Fig. 4B; Movie 5).

To determine the impact of progression of the cell cycle on SE fusion, we used roscovitine, which inhibits cyclin-dependent kinase (CDK) activity (Planchais et al., 1997). Pistils from pFWA::FWA-GFP were pollinated by the pRPS5A::H2B-tdTomato nuclear marker transgenic plant, and the ovules were analyzed by confocal microscopy in medium containing roscovitine. Out of 13 ovules, ten failed to enter mitosis, among which only three ovules displayed an influx of FWA-driven GFP into the persistent synergid (Fig. 4C,D; Movie 6). Thus, reduction of CDK activity by roscovitine likely induces the SE fusion defect as also does mitotic arrest at G2/M transition, which is earlier than the spindle checkpoint arrest induced by oryzalin during M-phase. These data indicate striking involvement of CDK-mediated M-phase entry in SE fusion. Although SE fusion could be a subsidiary event accompanying the G2/M transition, it is tempting to consider that the SE fusion-related factors are directly activated by the CDK–cyclin complex. Among seven CDK families in Arabidopsis, CDKA and CDKB are mainly implicated in cell cycle progression (Menges et al., 2005; Inzé and De Veylder, 2006), and roscovitine is known to inhibit both proteins (Nakai et al., 2006). An in vitro assay of protein phosphorylation with or without roscovitine may be a good approach to examine the regulatory mechanism of putative SE fusion factors in the future.

Taken together, we conclude that cytoskeletal actin and microtubules are not required for SE fusion. In contrast, CDK activity and proteins synthesized after fertilization are required in the endosperm for SE fusion. These observations have cleared the path for identification of the protein involved in this unique mode of plant cell fusion.

A DNA fragment of coding the first 30 amino acids of endo-xyloglucan transferase (EXGT-A1, At2g06850) followed by the CLOVER (Lam et al., 2012) was cloned into pENTR/D-TOPO vector (Invitrogen, Japan) to generate pOR039. LR recombination between the pOR039 and pSAN37 (Maruyama et al., 2015) was performed using LR clonase II (Invitrogen) to produce pDM377, a binary vector containing the pDD65::SP-CLOVER.

Col-0 was used as the wild-type plant. The pFWA::FWA-GFP, pFWA::ACT8, and pFWA::DN-ACTIN transgenic plants containing the pFWA::Lifeact-Venus marker, and the pFWA::Lifeact-Venus original line were described previously (Kinoshita et al., 2004; Kawashima et al., 2014; Kawashima and Berger, 2015). The pAGL80::tagRFP-TUA5 was kindly provided by Shuh-ichi Nishikawa (Niigata University, Japan). The pRPS5A::H2B-tdTomato was a gift from Daisuke Kurihara (Nagoya University, Japan; Adachi et al., 2011). The pDD65::SP-CLOVER was introduced into wild-type plant by the floral dip method (Clough and Bent, 1998). Plants were grown on soil at 22°C under continuous light.

To make stock solutions for the inhibitor assay, latrunculin A, oryzalin, BFA and cordycepin were dissolved into DMSO to 25 mM, 2.5 mM, 12.5 mM and 50 mM, respectively. We dissolved roscovitine into DMSO to 5 mM. The stock solution was added to the assay medium at a 250-fold dilution. Cycloheximide was dissolved in sterile water to prepare a 50 mg/ml stock solution. Those stock solutions were supplemented at a 250-fold dilution with the assay medium (half-strength Murashige and Skoog's medium, 5% sucrose, pH adjusted to 5.7 with KOH) before use. The assay media without supplements or with 0.4% DMSO were used as controls for the media containing cycloheximide or other inhibitors, respectively.

In an analysis of the pFWA::ACT8 and pFWA::DN-ACTIN transgenic plants containing the pFWA::Lifeact-Venus marker, pistils were pollinated with wild-type pollen, and the ovules were dissected into the assay medium at 14 hap. These ovules were transferred on a glass-bottom dish (D141410, Matsunami Glass Ind., Japan) and observed with a microscope (Ti-E, Nikon, Japan) equipped with a CSU-X1 disk-scan confocal system (Yokogawa, Japan), a CMOS camera (Hamamatsu Photonics, Japan) and a 515-nm laser line. Time-lapse imaging was performed as follows. Ovules were dissected at 6.5 to 8 hap from the pistils into the assay medium. The samples were excited with 488-nm and/or 561-nm irradiations, and confocal images were captured every 10 min using multiple z-planes (6–8 planes, 3.0-µm intervals) for 6 or 6.5 h. For the inhibitor assays with cycloheximide, latrunculin A and oryzalin, we used a disk-scan confocal system (CSU-XI, Yokogawa Electric, Tokyo, Japan) and an EM-CCD camera (Evolve 512, Photometrics, Tucson, USA). The other time-lapse imagings were performed with a spinning-disk confocal system (CellVoyager CV1000, Yokogawa Electric, Tokyo, Japan). In the analysis of the pDD65::SP-CLOVER transgenic plants, unfertilized ovules were dissected into the assay medium containing DMSO or BFA. The samples were excited with 950-nm irradiation, and the emissions between 500 and 550 nm, and 575 and 610 nm were collected at 2, 4 or 6 h after dissection using multiple z-planes (16 planes, 1.0-µm intervals) by a laser-scanning microscope (LSM780-DUO-NLO, Zeiss, Germany) equipped with 63× oil-immersion objective lens (Plan-APOCHROMAT, WD=0.19 mm, NA=1.40; Zeiss) and a Chameleon Vision II laser (Coherent, USA).

Fluorescent images with multiple z-planes were processed with CV1000 software (Yokogawa Electric) or ZEN2012 software (Zeiss) to create the maximum intensity projection images. ImageJ (http://imagej.nih.gov/ij/) and QuickTime Player 7 Ver. 7.7.1 were used for image-adjusted brightness and contrast and movie editing of the time-lapse analyses.

We thank Y. Sato for technical support for imaging and comments on the manuscript, S. Nishikawa for the pAGL80::tagRFP-TUA5 marker, D. Kurihara for pRPS5A::H2B-tdTomato, M. Z. Lin for pcDNA3-Clover, and T. Sasaki and H. Shikata for advising on chemical treatment. Most of this work was conducted at the Institute of Transformative Bio-Molecules (WPI-ITbM) at Nagoya University, supported by the Japan Advanced Plant Science Network.