The male gamete membrane protein DMP9/DAU2 is required for double fertilization in flowering plants

Flowering plants produce seeds through double fertilization, a unique manner of sexual reproduction in organisms. After pollination, a germinated pollen tube delivers a pair of sperm cells to female gametes (an egg and a central cell) that develop in an embryo sac. After the male and female gametes meet, one sperm cell fuses with the egg cell and another sperm cell fuses with the central cell to produce an embryo and an endosperm, respectively. The double fertilization system offers advantages to the conservation and birth of species in flowering plants. In addition, many crops have been modified by artificial cross-breeding based on the double fertilization phenomenon. Understanding the basic mechanism of double fertilization would contribute to the development of new breeding techniques.

Double fertilization proceeds through the recognition, attachment and membrane fusion of male and female gametes, and these processes are directly regulated by proteins on the gamete surface (Mori et al., 2015; Dresselhaus et al., 2016). Male gametic membrane proteins GCS1/HAP2 and GEX2 have been identified as fertilization regulators that directly regulate gamete fusion and attachment, respectively (Mori et al., 2006, 2014; von Besser et al., 2006). However, a comprehensive understanding of the entire molecular mechanism is lacking. To identify additional fertilization regulators, in this study, we performed the proteomic analysis of membrane proteins in L. longiflorum generative cells. One of the identified proteins is an ortholog of Arabidopsis DMP9/DAU2 (Kasaras and Kunze, 2010; Borg et al., 2011), the function of which remains unknown. We have characterized DMP9 as a double fertilization regulator that is involved in gamete interaction.

Microsomal fractions derived from mature pollen and isolated generative cells of L. longiflorum were subjected to proteomic analysis using the RNA sequence database of L. longiflorum generative cells (T.M., unpublished) to detect membrane-abundant proteins in the generative cells (Fig. S1A). As a result, two peptide sequences derived from an identical cDNA contig (k47_NODE_2124_cov_1342.396240_364nt, Fig. S1B) were significantly detected in the generative cells (Table S1). After isolation of the full-length sequence, the cDNA (GenBank Accession Number LC404162) was found to encode a homolog of Arabidopsis thaliana protein of unknown function DMP9/DAU2 (At5g39650) (Fig. S1C) (Kasaras and Kunze, 2010; Borg et al., 2011). DMP9 has been reported as a member of a family of membrane-localized proteins (DMP1-10), and DAU2 has been reported to be a sperm-specific gene regulated by the male gamete differentiation factor DUO POLLEN 1 (DUO1). Lily and Arabidopsis DMP9 proteins (designated as LlDMP9 and DMP9, respectively) are predicted to have four putative transmembrane domains and putative DMP9 orthologs have also detected as similar membrane proteins in other green plant species (Fig. S2A) (Kasaras and Kunze, 2010).

To analyze expression patterns, semi-quantitative RT-PCR was performed in lily samples. LlDMP9 was specifically expressed in mature pollen (Fig. 1A). The accumulation of LlDMP9 transcripts was initiated after generative cell formation by pollen mitosis I, which was completed at 70 mm bud length, and gradually increased as pollen development progressed (Fig. 1B). In situ hybridization revealed that the LlDMP9 transcripts specifically accumulated in the generative cell cytoplasm (Fig. 1C). In addition, RT-PCR analysis of DMP9 also showed pollen-specific expression, supporting a previous study (Fig. S3A) (Kasaras and Kunze, 2010). Next, we produced Arabidopsis transgenic plants expressing DMP9-GFP by introducing pDMP9::DMP9-GFP into the HTR10-RFP marker line, in which male germline nuclei are specifically labeled with RFP (Ingouff et al., 2007). Similar to the LlDMP9 expression pattern during pollen development, the DMP9-GFP signal appeared in generative cells and was also maintained in the sperm cells of mature pollen (Fig. 1D). The GFP signal likely localized to the sperm surface rather than to the endomembrane system (Fig. S3B). DMP9-GFP driven by the 35S cauliflower mosaic virus promoter showed plasma membrane localization in onion epidermal cells (Fig. S3E,F), suggesting that this localization is not controlled by male gamete-specific factors. In contrast to GCS1 and GEX2, which preferentially reside in the sperm endomembrane system before fertilization, DMP9 is distributed more to the plasma membrane (Fig. S3B-D). It has been reported that the distribution of GCS1/HAP2 shifts to the sperm cell surface upon secretion of EGG CELL 1 (EC1) from the egg cell, resulting in sperm cell activation and fusion with female gametes (Sprunck et al., 2012). Considering the default localization of DMP9 in sperm cells, it is suggested that DMP9 on the sperm surface functions prior to or simultaneously with GCS1/HAP2 and GEX2. A recent study reported that two isoforms of DMP1 (DMP1.1 and 1.2) showed different destinations, the tonoplast (TP) and the plasma membrane (PM), within a cell (Kasaras and Kunze, 2017). The fact that DMP9 has no signal sequence at the N terminus and shows no TP localization in onion epidermal cells also strongly suggests that DMP9 localization is restricted to the plasma membrane in sperm cells.

To assess DMP9 function, we produced DMP9 knockdown plants (DMP9^KD) by RNA interference (RNAi) because there are no available T-DNA insertion lines in the seed bank. The RNAi construct targeting DMP9 was introduced into a DMP9-GFP line where DMP9 was labeled with GFP in the sperm cells (Fig. 1D, Fig. S3B). Therefore, the disappearance of GFP from sperm cells in RNAi transgenic lines demonstrates the knockdown of DMP9 expression (Fig. S4A,C). In the obtained DMP9^KD lines, reduced seed sets were observed (Fig. S4B). Compared with the original DMP9-GFP line, aborted seed development was frequently observed in the DMP9^KD lines (7.8-21.5%), whereas the frequencies of undeveloped seeds were similar in all lines (Fig. 2A,B). In addition, reciprocal crosses between a wild-type line and a DMP9^KD line revealed that seed abortion occurred when DMP9^KD pollen was used, suggesting that DMP9 reduction causes functional abnormalities in male gametophytes (Fig. 2B).

A similar aborted seed phenotype has been reported in mutants that cause single fertilization of either of the female gametes (Ron et al., 2010; Mori et al., 2014). To address the cause of the aborted seed development, DMP9^KD sperm behavior during double fertilization was observed. DMP9^KD pollen was used to artificially pollinate the egg cell double marker (DM) line pDD45::GFP-PIP2a/pEC1::H2B-RFP in which the egg cell membrane and the nucleus are labeled with GFP and RFP, respectively (Igawa et al., 2013). As a result, it was frequently observed that the signal in one of the sperm nuclei diffused and moved to the central cell, whereas the other sperm nucleus was condensed and remained on the egg cell plasma membrane (Fig. S5A), suggesting that single fertilization of the central cell occurred to result in seed abortion.

To analyze fertilization patterns more precisely, the behaviors of sperm cells from DMP9^KD line 3 (Fig. 2B) in ovules were observed at 9-10 and 16-18 h after pollination (HAP). In the observation analysis of ovules (n=472 in total) at 9-10 HAP, dispersed and condensed sperm nuclear signal morphologies were categorized as markers of the fused and unfused states, respectively (Fig. 2C). In the case of double fertilization, two dispersed sperm RFP signals were observed in the ovule (Fig. 2C, left panel). As a result, 81.8% (386/472) of ovules showed successful double fertilization. Single fertilization, whereby one dispersed RFP signal and one condensed RFP signal were detected in the ovule, was also observed; the dispersed RFP signal on the chalazal or micropylar side of the ovule reflects fusion with the central cell or the egg cell, respectively (Fig. 2C, middle and right panels). In the case of single fertilization, the frequency of gamete fusion by DMP9^KD sperm cells was significantly higher for the central cell (83/472, 17.6%) than for the egg cell (3/472, 0.6%), whereas the frequencies in both types of single fertilization by wild-type sperm cells were similar (3/357 ovules, 0.8%) (Fig. 2D). It has been reported that there is a time lag of several minutes between the first fertilization and the second one (Hamamura et al., 2011). Even though the observed single fertilization patterns included such cases of time lag, it is still noteworthy that stagnant fertilization by DMP9^KD sperm cells is almost exclusive to the egg cell.

The DMP9^KD pollen was further used to fertilize another female DM line pFWA::FWA-GFP/pEC1::H2B-RFP, where the central cell and the egg cell nucleus were labeled with GFP and RFP, respectively (Mori et al., 2014). The success of central cell fertilization could be monitored on the basis of endosperm development showing multiple nuclear signals (Aw et al., 2010) (Fig. 2E, Fig. S5B). At 16-18 HAP, 81.3%±2.2% (mean±s.d.; n=9 pistils, n=337 ovules in total) of ovules showed endosperm development without remaining sperm nuclear signals, which was categorized as successful double fertilization. The frequency of non-fertilization, in which two sperm nuclei remained in an ovule without endosperm development, was 1.5±0.7%. In addition, 17.2±1.8% of ovules showed endosperm development with a single sperm remnant near the egg cell nucleus, reflecting single fertilization of the central cell (Fig. 2E, Fig. S5B). On the other hand, ovules containing a single sperm signal without endosperm development, which is expected in the case of single fertilization of the egg cell, were never observed. The frequencies of aborted seed development (16.9% in DMP9^KD line 3) (Fig. 2B) nearly corresponded to that of single fertilization of the central cell, strongly suggesting that the seed abortion in DMP9^KD plants was mainly caused by the absence of sperm-egg fusion. The behaviors of DMP9^KD sperm cells during double fertilization were also analyzed by time-lapse imaging of semi-in vitro-cultured ovules. As a result, we captured abnormal fertilization events in which no sperm-egg fusion or delayed sperm-egg fusion occurred (Movies 1 and 2). Based on these observations, we conclude that DMP9 functions as a fertilization regulator.

It has been experimentally proven that the two Arabidopsis sperm cells have no fertilization preference, based on the observation that both sperm cells have equal abilities to fuse with both female gametes, regardless of the order in which they are discharged from the pollen tube (Hamamura et al., 2011). Cytological studies of double fertilization by mutant sperm also suggest that each sperm cell freely chooses either of the female gametes (Chen et al., 2008; Ingouff et al., 2009; Ron et al., 2010; Maruyama et al., 2013; Kong et al., 2015). Despite the equal distribution of DMP9 in sperm cells, which is similar to that of GCS1/HAP2 and GEX2 (Fig. 1D, Fig. S3B-D), DMP9^KD obviously has more influence on egg cell fertilization. Therefore, it is considered that the two female gametes may exhibit differential sensitivity or reaction to DMP9.

To speculate on the role of DMP9 in double fertilization, we assessed the attachment of DMP9^KD sperm cells to the egg cell, as reported by Mori et al. (2014). Ovules with remaining sperm cell signals were treated enzymatically to detach the female gametes from adjacent cells in an embryo sac. As a result, 87.2% (34/39) of the ovules showed that single sperm remnants were associated with the plasmolyzed egg cell (Fig. 3A,D). The percentage was greater than that of the ovules containing a detached sperm cell (12.8%, 5/39 ovules) (Fig. 3B,D). In addition, we occasionally observed two sperm cell signals in an ovule, and 91.7% (11/12 ovules) of them showed attachment to the egg cell. The percentages of attached DMP9^KD sperm cells were similar to those of gcs1 mutant sperm cells (85.7%, observed in 18/21 ovules) (Fig. 3C,D) (Mori et al., 2014), reflecting that the unfused DMP9^KD sperm cell at least succeeded in attaching to the egg cell. Therefore, it is proposed that DMP9 is involved in a step that occurs during or after gamete attachment but before plasma membrane fusion.

LlDMP9 and DMP9 are specifically expressed in male gametes. The Arabidopsis ortholog of DMP9 has been reported to be activated by DUO1 (Borg et al., 2011), implying that DMP9 is expressed for sperm cell specification, along with GCS1/HAP2 and GEX2 (Brownfield and Twell, 2009). Based on the observations of DMP9^KD sperm cell behavior, we conclude that DMP9 is involved in fertilization control. DMP9 preferentially localizes to the plasma membrane soon after sperm cell differentiation, in contrast to the abundant localization of GCS1/HAP2 and GEX2 in the endomembrane system until fertilization. Knockdown of DMP9 causes seed abortion due to biased fertilization failure of the egg cell. Furthermore, unfused DMP9^KD sperm cells adhere to the egg cell. These findings strongly suggest the presence of a specific system regulating the sperm-egg interaction, and DMP9 may function during or after attachment to the egg cell. A similar idea has been proposed based on the preferential central-cell-only fertilization by a single sperm-like cell that was produced by diphtheria toxin A (DTA) expression driven by the GCS1/HAP2 promoter (Frank and Johnson, 2009). The perturbed translation of DMP9 by DTA could have resulted in the preferential fertilization failure of the egg cell in the previous study. Another study reported that the discharged sperm cells are rectified until two fertilization pairs are established (Huang et al., 2015). Huang et al. suggested the presence of a system in which each female gamete verifies the proper adhesion of the sperm cells. Considering our results and the previous findings, it may be that DMP9 is involved in the signaling that helps an egg cell correctly recognize attachment to a single sperm cell. The strictness of this recognition system may be regulated by the amount of DMP9 involved in the reaction. The delay in egg cell fertilization observed in time-lapse imaging (Movie 2) may support this idea. DMP9 is conserved in green plants, including species that perform egg-sperm fertilization in sexual reproduction (Fig. S2A). A previous study of glauce (glc) suggested that each female gamete develops competence for fertilization independently (Leshem et al., 2012). It is possible that gamete recognition based on DMP9 is conserved among green plant lineages as a mechanism specific to sperm-egg interactions, but not conserved in sperm-central cell interactions that may have evolved independently after the emergence of flowering plants. Interestingly, a previous study has showed that the expression of putative Chlamydomonas DMP9 ortholog was induced in both plus and minus types of gametes upon gamete activation (Ning et al., 2013) (Fig. S2B). Future comparative analysis of DMP9 between different green plant taxa may provide additional insights into the evolution of plant sexual reproduction systems.

This study reports for the first time, the preferential fertilization block shown by two equally mutated sperm cells. Further analysis of DMP9 protein function and the relationship between DMP9 and known fertilization regulators (Mori et al., 2006, 2014; von Besser et al., 2006; Sprunck et al., 2012) will provide more insights into the molecular mechanism of double fertilization control. Such analyses will help unravel the mystery of how two independent male-female pairs are correctly established for successful double fertilization.

Trumpet lily (L. longiflorum. cv. Georgia) was grown in a greenhouse. The various stages of pollen development in this lily can be estimated from the length of the buds (Tanaka et al., 1987). Under our greenhouse conditions, the first pollen mitosis occurs in buds that are approximately 65 mm in length, and anthesis occurs when the buds reach 170 mm in length. We obtained microspores from buds that were 40 or 60 mm in length, bicellular pollen grains from buds that were 70, 90, 110, 130, 150 or 170 mm in length, and mature pollen grains from flowers 3 days after anthesis.

The background of all Arabidopsis transgenic marker lines was Col-0. Surface-sterilized seeds were germinated on Murashige and Skoog medium (pH 5.8) supplemented with 1% sucrose, 0.5 g/l 2-morpholino ethanesulfonic acid (monohydrate) and 0.8% agarose. pDD45::GFP-PIP2a/pEC1::H2B-RFP (Igawa et al., 2013) and pFWA::FWA-GFP/pEC1::H2B-RFP (Mori et al., 2014) lines were used as the female double marker (DM). DMP9-GFP and +/gcs1 expressing snRFP lines (Mori et al., 2014) were selected using 50 mg/l kanamycin. DMP9^KD lines were selected with 50 mg/l gentamycin. Seedlings were grown for 2-3 weeks and then acclimatized in soil. Plants were grown at 22°C under a 16-h light/8-h dark cycle.

Nearly mature pollen was collected from buds 150 mm in length. The pollen grains were washed in 0.5 M sucrose solution to remove surface oil. The washed pollen was applied to protein extraction in buffer containing 10 mM Tris-HCl (pH 6.8), 250 mM sucrose, 0.5 mM EDTA and 0.5% (w/v) Complete Mini protein inhibitor (Roche). After centrifugation of the initial extract at 20,000 g, the resulting supernatant was recovered. The supernatant was ultracentrifuged at 100,000 g to precipitate microsomal components. After removal of the supernatant, the microsomal fraction was extracted in EzApply sample buffer (ATTO). The generative cells were isolated from the same pollen sample, as previously described (Tanaka, 1988). The isolated generative cells were treated similarly as described above.

The microsomal proteins (10 μg) were subjected to SDS-PAGE. The gel was silver stained using an EzStain Silver kit (ATTO) according to the manufacturer's instructions. LC-MS/MS analysis of silver-stained proteins was performed as previously described (Igawa et al., 2017). For MASCOT analysis of each peptide, RNA sequencing of L. longiflorum generative cells was performed (InfoBio).

Based on the contig sequences of LlDMP9, 5′ and 3′ RACE-PCRs were carried out using a GeneRacer kit (ThermoFisher Scientific), with the primers LlDMP9f1, LlDMP9f2, LlDMP9r1 and LlDMP9r2 (Table S2). Each PCR product was cloned into the pCR-Blunt II-TOPO vector (ThermoFisher Scientific) and sequenced (Eurofins). The LlDMP9 accession number is LC404162.

For lily, RNA extraction from pollen samples from each developmental stage and cDNA production were performed as previously described (Ueda et al., 2000). For Arabidopsis, total RNA was reverse transcribed with SuperScriptIII reverse transcriptase (ThermoFisher Scientific) and oligo(dT) primers to synthesize first-strand cDNA. The cDNA samples were subjected to PCR using primer sets of LlDMP9f1 - LlDMP9r1 and LlEFf - LlEFr for LlDMP9, or primer sets of AtDMP9f7 - AtDMP9r7 and AtACT3f - AtACT3r for DMP9 (Table S2).

For the production of RNA probes, the full length of LlDMP9 cDNA was amplified using the primer set DMP9f(Sac) and DMP9r(Kpn) (Table S2) and cloned into the pSPT19 vector (Roche) after digestion with KpnI and SacI. In vitro transcription for labeling of LlDMP9 was performed using a DIG RNA Labeling Kit (SP6/T7) (Roche) according to the manufacturer's instructions. The procedures and conditions for in situ hybridization on mature L. longiflorum pollen have been previously described (Ueda et al., 2000). Counterstaining was performed with 0.2% Methyl Green (Merck).

The DMP9 open reading frame (ORF) and the flanking 5′ untranslated region (UTR) (1.5 kb) were amplified with the primers AtDMP9f2(Bam) and AtDMP9r2(Xho) to add BamHI and XhoI recognition sites to the 5′ and 3′ ends of the amplicon (Table S2). The resulting fragment was cloned into the pCR-Blunt II-TOPO vector. After sequencing, the BamHI-XhoI fragment was excised and ligated into the pENTR 2B vector. The 5′UTR-ORF region of DMP9 was finally transferred to the pGWB4 (Nakagawa et al., 2007) through the LR reaction. The obtained pDMP9::DMP9-GFP construct was introduced into the HTR10-mRFP sperm nucleus marker line (Ingouff et al., 2007) by Agrobacterium infection.

The coding sequence of DMP9 lacking the stop codon was amplified using the primer AtDMP9f1 in combination with AtDMP9r1 from mature pollen cDNA. The amplicon was cloned into the pCR8/GW/TOPO vector (ThermoFisher Scientific) and sequenced. After LR reaction with pGWB4 (Nakagawa et al., 2007), p35S::DMP9-GFP was produced. The obtained construct was introduced into onion epidermal cells by particle bombardment as described previously (Mori et al., 2014).

The DMP9 promoter sequence (∼1 kb upstream of DMP9), DMP9-coding sequence, GUS linker and NOS terminator were amplified with the primer pairs AtDMP9promf(Kpn)-AtDMP9promr(Bam), AtDMP9f5(Xba)-AtDMP9r5(Bam), AtDMP9f6(Sal)-AtDMP9r6(Pst), GUSf(Xba)-GUSr(Sal) and NOStermf(Pst)-NOStermr(Hin) (Table S2), respectively, adding appropriate restriction enzyme recognition sites. These amplicons were cloned into the pCR-Blunt II-TOPO and then sequenced. Each of the amplicons was excised with suitable restriction enzymes and was ligated into the pPZP221 vector (Hajdukiewicz et al., 1994). The pDMP9::DMP9-RNAi construct obtained was introduced into an DMP9-GFP line by Agrobacterium infection.

The pollen at various developmental stages was stained with DAPI staining solution as reported by Park et al. (1998) and observed with an epifluorescence microscope (BX51, Olympus). To observe the sperm cells in a pollen tube, a wild-type pistil was pollinated with pollen from DMP9-GFP, GEX2-GFP (Mori et al., 2014) or GCS1-GFP plants (Igawa et al., 2013) (all lines carried the sperm nucleus marker HTR10-mRFP) and cut at the boundary of the style and the ovary. The cut end of the upper part of the pistil was half-embedded into thin (less than 1 mm) pollen tube medium (Boavida and McCormick, 2007) spread on a cover slip. The pistil was cultured at 22°C until the sperm cells observed in the pollen tube passed through the cut end. Images were captured with an Olympus DP-72 digital camera and merged using Adobe Photoshop CS5.

Total RNA was extracted from mature pollen of wild-type plants and each DMP9^KD line. After treatment with DNase, total RNA was reverse transcribed with SuperScriptIII reverse transcriptase and oligo(dT) primers to synthesize first-strand cDNA. Real-time PCR reactions were performed using KOD SYBR qPCR Mix (Toyobo, Osaka, Japan) and an Applied Biosystems StepOnePlus Real-Time PCR System. The target sequence of DMP9 was amplified with the primers AtDMP9f8 and AtDMP9r8 (Table S2). eIFG4 (At3g60240) as an internal standard was also amplified with the primers AteIFG4f and AteIFG4r (Table S2).

Three weeks after pollination, siliques were dissected under an SZX9 stereomicroscope (Olympus) to count the numbers of fruited, aborted and undeveloped seeds. For observation of DMP9^KD sperm behavior, female DM lines were emasculated before anthesis. Half a day later, the pistils were pollinated with excess DMP9^KD pollen. The ovules were excised from the pistil 10-18 HAP and were observed using a BX51 epifluorescence microscope.

Semi-in vitro double fertilization was performed as described previously (Hamamura et al., 2011), except that the pollen germination medium reported by Boavida and McCormick (2007) was used. At 5-6 h after pollination, double fertilization was observed via microscope using a previously described setting (Maruyama et al., 2013). Time-lapse images were acquired every 5 min with seven planes at intervals of 3.0 μm. Images were processed using Metamorph (version 7.8.1.0.0) (Universal Imaging). ImageJ software (rsb.info.nih.gov/ij/) was used to adjust the images and edit the movies.

An egg cell DM line pDD45::GFP-PIP2a/pEC1::H2B-RFP (Igawa et al., 2013) was used as the female parent. Pistils emasculated half a day before experiments were hand-pollinated with DMP9^KD or +/gcs1 pollen. At 14-15 HAP, the ovules were dissected and transferred to polysaccharide-digesting enzyme solution (Mori et al., 2014) supplemented with 2 mM CaNO₃ and 0.01% Silwet L-77. After incubation for ∼2 h at 32°C, the ovules were observed under a BX51 epifluorescence microscope.

For analysis of seed development patterns, the frequencies of each developmental pattern per silique were calculated, and at least nine siliques were sampled for each DMP9^KD line. Statistical significance between egg cell and central cell single fertilization was evaluated by chi-square test (Fig. 2D). The data shown in Fig. 2E are summarized as the frequencies of each fertilization pattern per pistil.

We thank Dr Ichiro Tanaka (Yokohama City University, Japan), for growing L. longiflorum used for isolation of generative cells and Dr Nakagawa (Shimane University, Japan) for pGWB4 vector.